Nitrogen Identified as cause of Salt Marsh Loss

LETTER doi:10.1038/nature11533

Coastal eutrophication as a driver of salt marsh loss

Linda A. Deegan1, David Samuel Johnson1,2, R. Scott Warren3, Bruce J. Peterson1, John W. Fleeger4, Sergio Fagherazzi5

& Wilfred M.Wollheim6

Salt marshes are highly productive coastal wetlands that provide

important ecosystem services such as storm protection for coastal

cities, nutrient removal and carbon sequestration. Despite protective

measures, however, worldwide losses of these ecosystems have

accelerated in recent decades1. Here we present data from a nineyear

whole-ecosystem nutrient-enrichment experiment. Our study

demonstrates that nutrient enrichment, a global problem for

coastal ecosystems2–4, can be a driver of salt marsh loss. We show

that nutrient levels commonly associated with coastal eutrophication

increased above-ground leaf biomass, decreased the dense,

below-ground biomass of bank-stabilizing roots, and increased

microbial decomposition of organic matter. Alterations in these

key ecosystem properties reduced geomorphic stability, resulting

in creek-bank collapse with significant areas of creek-bank marsh

converted to unvegetated mud. This pattern of marsh loss parallels

observations for anthropogenically nutrient-enriched marshes

worldwide, with creek-edge and bay-edge marsh evolving into mudflats

and wider creeks5–7. Our work suggests that current nutrient

loading rates to many coastal ecosystems have overwhelmed the

capacity of marshes to remove nitrogen without deleterious effects.

Projected increases in nitrogen flux to the coast, related to increased

fertilizer use required to feed an expanding human population,may

rapidly result in a coastal landscape with less marsh, which would

reduce the capacity of coastal regions to provide important ecological

and economic services.

An accelerated global nitrogen cycle1–3 has greatly increased the flow

of reactive nitrogen (primarily as NO3

2) from land to coastal marine

ecosystems, causing harmful algal blooms, hypoxia and fisheries

losses5,8. Salt marshes occupy a critical interface between the land

and the sea, where they provide important ecological and economic

services, such as nutrient removal, storm protection for coastal cities

and carbon sequestration, and habitats for numerous species of fish,

birds and invertebrates. It is thought that salt marshes can protect

coastal bays by removing land-derived nutrients9,10, a conclusion based

on measures of whole-system nutrient budgets11,12 and plot-level

experiments in which added nutrients were transformed into greater

above-ground plant production (primarily cordgrass Spartina spp.) or

denitrified4. Globally between a quarter and half of the area of the

world’s tidal marshes has been lost, and although multiple factors (sealevel

rise, development, loss of sediment supply) are known to contribute

to marsh loss1, in some locations the drivers remain unexplained.

Understanding the mechanisms underlying the continued loss of this

ecologically and economically important ecosystem is a global priority.

Here we present an ecosystem-level experimental approach to

understanding how the intertwined responses of plant biomass allocation,

microbial decomposition, and geomorphic stability to coastal

nutrient enrichment may drive salt marsh loss. For nine years (2004–

2012) we have enriched multiple whole-ecosystemmarsh landscapes to

nutrient levels that correspond to moderately-to-highly eutrophic waters

by adding dissolved nutrients to flooding tidal water13. Approximately

50%–60% of the added NO3

2 was processed (assimilated or denitrified)

in the nutrient-enriched systems; the remainder was exported in

ebbing tidal water12. The large scale of this experiment, which included

creeks, mudflats, tall-form smooth cordgrass (Spartina alterniflora) at

the creek-channel edge and saltmeadow cordgrass (S. patens) in the

high marsh, has revealed interactions that would not be apparent from

plot-level experiments in individual habitats.

Nutrient enrichment may invoke a series of positive feedbacks by

altering ecosystem processes that affect below-ground dynamics and

creek-bank stability, leaving marshes more susceptible to the erosive

forces of storms and sea-level rise and gravitational slumping. In less

than a decade, a cascade of changes induced by nutrient enrichment

resulted in loss of low marsh along the creek-bank edge (Fig. 1a–f) and

a corresponding loss of ecosystem function. Smooth cordgrass along

the creek-bank edge responded to nutrient enrichment with increased

above-ground biomass expressed as heavier, taller shoots (Fig. 2a),

lower structural compounds (decrease of about half in foliar lignin),

and increased N content (Table 1), with response ratios comparable to

plot-level nutrient-enrichment experiments4,14. Increased plant height

coupledwith less structural tissue causedmore extensive areas of smooth

cordgrass to fall over (lodge)—awell-known response to over-fertilizing

grasses15.Using permanent transects and high-precision global positioning

system (GPS) mapping across the elevation gradient, we found no

evidence (D.S.J., R.S.W. and L.A.D., manuscript in preparation) for the

hypothesized shift in the up-elevation boundary between S. alterniflora

and S. patens in response to nutrients15. In nutrient-enriched marshes,

smooth cordgrass allocated less photosynthate to nutrient-gathering

roots and storage rhizomes, resulting in a third less total below-ground

biomass and a lower root:shoot ratio (Fig. 2b, c). Two smooth cordgrass

growth attributes, a highly plastic above-ground/below-ground allocation16

and foliar uptake ofNO3

2(ref. 17), contribute to the reductions in

total below-ground biomass observed in nutrient-enriched marshes.

The continuous availability of high NO3

2 in the water and more

decomposable marsh grass detritus (due to higherNcontent and lower

lignin) increased decomposition rates (Table 1). Whole-ecosystem

nitrate removal was 40 times higher in the nutrient-enriched marsh

and was primarily attributable to microbial use of the added NO3

2 to

decompose organic matter12. Potential denitrification—an indicator of

anaerobic microbial decomposition using nitrate as an electron

acceptor with the end product being N2 gas—increased 1.7-fold in

creek bank sediments, while litter respiration—a measure of aerobic

microbial decomposition—almost doubled (1.9-fold). Denitrification

is the highest energy-yielding decomposition process in anoxic marsh

sediments and is favoured in the presence of high nitrate5. Accelerated

decomposition increased the fraction of fine detrital organic matter,

with 65% of the cores from nutrient-enriched creeks having a higher

percentage of fine organic matter. As a result, the fine-grained, lessconsolidated

creek banks retained more water at low tide (Fig. 2d).

The combination of fewer roots and rhizomes, drag by tidal currents

on lodged plants, more decomposed organic matter and higher water

content undermines the structural integrity of the creek bank such that

the effects of standard physical forces become enhanced. Loss of roots

1The Ecosystems Center, Marine Biological Laboratory, 7 MBL Street, Woods Hole, Massachusetts 02543, USA. 2Department of Biology, Sewanee University of the South, 735 University Avenue, Sewanee,

Tennessee 37383, USA. 3Department of Botany, Connecticut College, 270 Mohegan Avenue, New London, Connecticut 06320, USA. 4Department of Biological Sciences, Louisiana State University, Baton

Rouge, Louisiana 70803, USA. 5Department of Earth and Environment, Boston University, 675 Commonwealth Avenue, Boston, Massachusetts 02215, USA. 6Department of Natural Resources and the

Environment, University of New Hampshire, 8 College Road, Durham, New Hampshire 03824, USA.

388 | NATURE | VOL 490 | 18 OCTOBER 2012

©2012 Macmillan Publishers Limited. All rights reserved

and rhizomes (which bind sediments and provide drainage macropores)

and loss of large organic matter particles (which form air pockets

that can help drain creek banks) both contribute to increased creekbank

water content18. The higher pore water pressure in the bank

reduces the frictional shear strength of the soil and increases the sliding

force by adding weight to the creek bank19. At low water, the weight of

the saturated bank exceeds the cohesive forces holding it together, the

top of the bank cracks and creek-bank sections slide downward by

gravitational slumping18. The structural failure of the creek edge

implies that tidal forces, which under non-eutrophic conditions can

be withstood, overcome the lowered cohesive strength of the nutrientenriched

bank habitat. Cracks developed over time with nutrient

enrichment (Fig. 2d), and after seven years of enrichment, there were

more (threefold) and longer (4.5-fold) fractures at the top of the bank

parallel to the creek (Table 1) and large blocks of low marsh slumped

into the creek (Fig. 2e). Without the buttressing edge of low marsh,

high-marsh turf sheared from the sediment at the base of the active

rooting layer creating ‘toupees’ that slid down slope into creek channels

(Supplementary Fig. 1), tripling the area of bare mud over time

(Table 1). The average width of the band of tall S. alterniflora along

the creek edge decreased from approximately 3m to approximately

2m wide and became highly reticulated (Fig. 1c versus Fig. 1f; Fig. 2e)

in nutrient-enriched systems. High-resolution measurements of

channel cross-sections between the fifth and eighth years of enrichment

indicate that the maximum erosion rate of creek banks was

0.2m3 per year per metre of channel length. Loss of marsh along creek

channel edges contrasts sharply with models that suggest that higher

above-ground plant biomass in response to nutrients would in turn

trap more sediment and stabilize marsh edges relative to sea-level

rise20. Slumping and cracking leads to a positive feedback, with

increased infiltration by nutrient-rich water into sediments, which

stimulate microbial decomposition of peat and further weakens sediments.

As nutrient-enriched creek banks collapse and retreat, channels

widen, increasing the unvegetated intertidal area at the expense of

vegetated marsh.

The generality of our whole-ecosystem experiment is supported by

extensive process work in small marsh fertilization plots and in anthropogenically

nutrient-enriched estuaries. From Louisiana to Nova

Scotia, nutrient enrichment has been shown consistently to increase

above-ground plant biomass14,16, mainly to decrease but sometimes

not to change below-ground biomass16,21–24 and to increase decomposition22,25.

The decomposition response might be expected to be

stronger inNorthernmarshes with a high percentage of organic matter,

but because the organic content of our creek-bank marsh is in the

middle of the range for Atlantic coast marshes (Methods), we expect

our decomposition response is typical. The combination of root loss

and increased decomposition can decrease soil strength26. The development

of cracks that lead to marsh loss has been approximately linear

over the initial nine-years; however, this is a process that will play out

over decades. Much uncertainty remains about whether this process

of marsh loss is self-limiting, because creek banks may eventually

be stabilized by the decrease in slope steepness from slumps, or selfreinforcing

owing to fractures facilitating the seepage of nutrientenriched

water into the bank and stimulating decomposition. Other

local environmental factors (such as tidal range, temperature and sediment

deposition) may modify the effects of nutrients on ecosystem

processes. Therefore, understanding the whole-ecosystem response to

nutrient loading across broad environmental gradients and longer

timescales requires more experimental ecosystem-level studies.

Many salt marshes may be at risk, because nutrient fluxes to the

coast have increased worldwide, with the largest increases in N flux

occurring at coastlines with large areas of intertidal marshland in the

temperate zones of eastern North America, Europe and eastern China

(Fig. 3a). A recent survey27 documented 415 eutrophic coastal systems

of concern worldwide, and found that only 13 systems were in recovery.

There is evidence that salt-marsh loss in Europe5 and along the Atlantic

coast of the United States 6,7 may be driven, in part, by anthropogenic

nutrient enrichment. Along Long Island Sound, Connecticut, USA,

coincident with increased total N in runoff, several marshes lost 27%–

54% of their low marsh (Fig. 3b), but very little high marsh, resulting





Edge: channel

ratio = 1.1

Edge: channel

ratio = 1.5

25 m

25 m

a b

d 2010 e 2010

c 2010

f 2010

Figure 1 | Comparison photos of the marshes from the ecosystem nutrient-enrichment experiment. a–c, Reference. d–f, Nutrient-enriched. Photo credits:

a, b, d and e, L.A.D.; c and f, Google Earth (19 June 2010 image, copyright 2012 Google).


1 8 O C T O B E R 2 0 1 2 | VO L 4 9 0 | N AT U R E | 3 8 9

©2012 Macmillan Publishers Limited. All rights reserved

1.5 1.5











1 2 3

0 20 40




Shoot height

Below-ground live biomass

(g m–2 cm–1)


Years of nutrient enrichment

Reference Nutrientenriched

Root:shoot ratio

Response ratio










Depth (cm)

Depth (cm)

Fracture density in bank

Number of slumps in channel

5 6 7

0 1 2 3


4 5

Years of nutrient enrichment

6 7

53 55 58 60










Large Small Total





a b c

d Creek-bank water content (%) e f

Figure 2 | Ecosystem attributes of reference and nutrient-enriched salt

marshes after up to 7 years of nutrient enrichment. a–c, Plant attributes.

a, Annual response ratio (nutrient-enriched/reference) for above-ground shoot

attributes. Responses were comparable to other marsh nutrient-enrichment

studies along the North American Atlantic coast and, as is typical in marshes,

the strength was variable from year to year (green indicates average values of

fertilized marshes greater than reference marshes, n56 per year; solid green

line is shoot specific mass; dashed green line is shoot height; points above the

response-ratio of 1.0 (horizontal black line) indicate increased growth in

response to fertilization in that year). b, Below-ground root and rhizome

biomass in reference (open circles) and nutrient-enriched (solid green squares)

marshes (2010; n520 per treatment). c, The above-ground to below-ground

plant biomass ratio (n520 per treatment). d–f, Creek-bank geotechnical and

geomorphic attributes. d, Vertical profile of percentage water content in

reference (open circles) and nutrient-enriched (solid green squares) marsh

creek-bank sediments with depth (2010; n520 per treatment). e, Fracture

density in high marsh plotted against years of nutrient enrichment. Annual

means are shown for reference (open squares), the start of enrichment in 2009

(solid squares) and the start of enrichment in 2004 (solid circles). f, Number of

low-marsh creek-bank slumps in the channel in reference (open bars) and

nutrient-enriched (solid green bars) marshes (2010; n52). Values are all

mean6standard error.

Table 1 | Response of salt marsh ecosystem properties to chronic nutrient enrichment

Ecosystem properties Response ratio Mean (standard error) P-value

Reference Nutrient

Vascular plants

Shoot height (cm)* 1.1 146 (3) 154 (2) 0.04

Shoot mass (g)* 1.2 6.18 (0.28) 7.18 (0.18) 0.01

Shoot specific mass (g cm21)* 1.1 0.042 (0.002) 0.046 (0.001) 0.02

Lodging (%) 0 (0) 41 (2) ,0.001

Foliar N (%)* 1.1 1.43 (0.08) 1.54 (0.09) ,0.001

Lignin (%)*** 0.6 30 (7.5) 17 (1.6) 0.07

Below-ground biomass (gm22) 0.7 579 (60) 387 (64) 0.08

Creek geomorphology and sediment geotechnical properties

Fracture density (number per 50m of creek edge)** 3 1.1 (0.2) 3.3 (0.7) ,0.001

Amount of creek bank with fractures (%)** 4.5 6.6 (0.7) 29.5 (2.6) ,0.001

Fracture length (m)** 4.5 3.3 (0.4) 14.7 (1.7) 0.002

Exposed mud area (%)** 3 7.5 (1.7) 22.8 (4.6) 0.004

Total number of slumps in channel per creek 2.1 19.0 (1.0) 40.5 (11.5) 0.07

Fine organic matter (%) 1.7 16.1 (1.9) 26.7 (4.5) 0.17

Water content (%) 1.04 56.2 (0.4) 58.5 (0.1) 0.02

Channel width/depth ratio 1.3 3.02 (0.11) 3.86 (0.02) ,0.001

Microbial decomposition processes

Plant litter respiration

(CO2 g21 S21)

1.9 1.29 (0.12) 2.49 (0.33) 0.04

Potential denitrification in creekbank sediment

(nmol g21 h21)***

1.7 40.7 (7.9) 70.8 (6.3) 0.01

A response ratio (nutrient-enrichment/reference) greater than one indicates a positive response to nutrient enrichment. P-values indicate the effect of nutrient enrichment on response variables; see ‘Statistical

summary’ in Methods. Means (6s.e.m., standard error of the mean) were calculated from the data averaged by creek each year (N52 for nutrient-enriched and N52 for reference, except where noted). *Means

(6s.e.m.) calculated from data averaged by creek and pooled across seven years of data. **Means (6s.e.m.) calculated from long-term nutrient-enriched and reference data averaged by creek and pooled across

two years (2009 and 2010) of data. ***Data averaged across subplots within creeks (N51 per treatment).


390 | NATURE | VOL 490 | 18 OCTOBER 2012

©2012 Macmillan Publishers Limited. All rights reserved

in wider creeks and increased mudflat area over the last three decades7.

In Jamaica Bay, New York, USA, the rate of creek and open-bay-edge

marsh loss in the 1990s was almost double that of the previous four

decades (134km2 yr21 versus 72km2 yr21) and the timing corresponds

to an increase in total nitrogen loading (up to 80 mM open-bay total

dissolved N) from sewage inputs6. Jamaica Bay marshes also have

reduced root and rhizome mass and more degraded sediment24, similar

to our experimentally nutrient-enriched marshes.

The loss of creek-channel marsh banks is particularly significant

because it is the most productive marsh zone and a critical area for

regulating (C storage and N removal) and provisioning (fisheries) ecosystem

services. Below-ground plant productivity disproportionately

contributes to detrital organic matter in a wide variety of ecosystems28,

suggesting that the measured reduction in below-ground allocation

may decrease the C storage potential of marshes. Marshes along creek

channel edges are hotspots for denitrification with rates of N removal

fivefold to tenfold higher than mudflats29, suggesting that marsh conversion

to mudflat decreases the ability of the coastal landscape to

remove N from tidal waters. Considerable research shows the importance

of the creek marsh edges in providing shelter and food to juvenile

fish and shrimp30,31, so the loss of marsh-edge habitat may ultimately

negatively affect fisheries production.

Simultaneous increases in nutrient loading and in sea-level rise may

result in synergistic marsh loss greater than the effects of either stressor

alone. For example, the higher wave energy and flow velocities associated

with sea level rise32 when combined with decreased creek-bank

stability induced by nutrient enrichment may accelerate erosion and

creek-bank loss. The drowning of high marsh due to sea-level rise and

loss of creek-edge marsh due to eutrophication, especially when exacerbated

by upland development that limits the ability of marshes to

move inland,may lead to a coastal landscapewith a dramatically reduced

capacity to provide important ecological and economic services.Nflux

to the coastal zone has already increased at least tenfold over preindustrial

levels and is projected to continue to increase as we ramp

up fertilizer use to produce food for the expanding human population

over the next few decades1,27. The potential deterioration of coastal

marshes due to eutrophication adds an unanticipated dimension to

the challenge of managing nitrogen while meeting food production in

the twenty-first century.


We enriched primary tidal creeks in Plum Island Estuary, Massachusetts, USA, to

nutrient levels corresponding to moderately-to-highly eutrophic coastal waters by

addingNand P to the twice-daily flooding tides for nine years (2004–2012) during

the growing season (about 120 days, 15 May–15 September), enriching about

30,000m2 of marsh per experimental primary creek system (N52 enrichment

started in 2004, N51 started in 2009, reference N52–6)13. Initial measurements

(1998–2003) found few differencesamong tidal creeks13, and other potential drivers

did not differ among treatmentmarshes or do not occur in the PlumIsland Estuary

(Supplementary Information).

To detect changes in plant biomass allocation, the height, dry weight, and

quality (percentage N, lignin content) of Spartina alterniflora above-ground

shoots were measured and below-ground cores were analysed for live roots and

rhizome biomass. Cores were also analysed for sediment geotechnical properties

(water content, percentage organic matter and particle size). To determine

changes in creek geomorphology, fractures in the vegetated marsh platform were

enumerated along 250–300m of creek banks and point-intercept transects indicated

the presence or absence of vegetation in the creek bank. Creek-bank blocks

that had slumped into tidal creeks were enumerated and creek width, depth and

erosion measured over time using a ‘total station’ and high-precision GPS surveys.

Microbial decomposition was measured as potential denitrification in the creek

bank and microbial respiration of surface litter.

Full Methods and any associated references are available in the online version of

the paper.

Received 16 April; accepted 20 August 2012.

1. Millennium Ecosystem Assessment. Ecosystems and Human Well-Being—

Synthesis Report (World Resources Institute, 2005).

2. Galloway, J. et al. Transformation of the nitrogen cycle: recent trends, questions

and potential solutions. Science 320, 889–892 (2008).

3. Canfield, D. E., Glazer, A. N. & Falkowski, P. G. The evolution and future of Earth’s

nitrogen cycle. Science 330, 192–196 (2010).

4. Verhoeven, J. T. et al. Regional and global concerns over wetlands and water

quality. Trends Ecol. Evol. 21, 96–103 (2006).

5. MacGarvin, M.Out of Sight,Out of Mind:Marine Eutrophication in the United Kingdom

(WWF-UK, 2001).

6. Hartig, E. K., Gornitz, V., Kolker, A., Mushacke, F. & Fallon, D. Anthropogenic and

climate-change impacts on salt marshes of Jamaica Bay, New York City. Wetlands

22, 71–89 (2002).

7. Tiner, R. W., Huber, I. J., Nuerminger, T. & Marshall, E. Salt Marsh Trends in Selected

Estuaries of Southwestern Connecticut. NWI Cooperative Report (US Fish and

Wildlife Service, Long Island Studies Program, Connecticut Department of

Environmental Protection, 2006).

Global change in dissolved-inorganic-nutrient

fluxes from pre-industrial to contemporary period

(kg km–2 yr–1)



< –200

–200 to –100

–100 to 0


0 to 100

100 to 200

200 to 500

> 500

Salt marsh






1970 1980 1990 2000


Proportion of

area in1974



Figure 3 | The global relationship between nutrient loading and salt-marsh

distribution and loss. a, The spatial distribution of the ramping up of

anthropogenic nitrogen loading (dissolved inorganic nutrient (DIN) fluxes)

fromcontinents to coastal oceans fromthe pre-industrial period (1800s) to the

contemporary period compared to global locations of salt marshes.

b, Conversion of low marsh to mudflat in six nutrient-enriched Long Island

Sound estuaries10 (see box in a indicating location of b). Substantial loss of low

marsh (solid line), smaller loss ofhighmarsh (dotted line) and increase inmudflat

(dashed line) area over time correlates with increased nutrients from sewage

treatment plants and runoff from land. Values are mean 6 standard error.


1 8 O C T O B E R 2 0 1 2 | VO L 4 9 0 | N AT U R E | 3 9 1

©2012 Macmillan Publishers Limited. All rights reserved

8. Diaz, R. J. & Rosenberg, R. Spreading dead zones and consequences for marine

ecosystems. Science 321, 926–929 (2008).

9. Valiela, I. & Cole, M. L. Comparative evidence that salt marshes and mangroves

may protect seagrass meadows from land-derived nitrogen loads. Ecosystems 5,

92–102 (2002).

10. Sousa, A., Lillebø, A. I., Caçador, I. & Pardal, M. Contribution of Spartina maritima to

the reduction of eutrophication in estuarine systems. Environ. Pollut. 156,

628–635 (2008).

11. Childer, D., Day, J. W. Jr & Mckeller, H. H. Jr. in Concepts and Controversies in Tidal

Marsh Ecology (eds Weinstein, M. & Kreeger, D.) 391–424 (Kluwer Academic,


12. Drake, D. et al. Salt marsh ecosystem biogeochemical responses to nutrient

enrichment: a paired 15N tracer study. Ecology 90, 2535–2546 (2009).

13. Deegan, L. A. et al. Susceptibility of salt marshes to nutrient enrichment and

predator removal. Ecol. Appl. 17, S42–S63 (2007).

14. Bertness, M. D. & Pennings, S. in Concepts and Controversies in Tidal Marsh Ecology

(eds Weinstein, M. & Kreeger, D.) 39–58 (Kluwer Academic, 2000).

15. Pinthus, M. J. Lodging in wheat, barley and oats: the phenomenon, its causes and

preventative measures. Adv. Agron. 25, 209–263 (1974).

16. Darby, F. A.&Turner, R. E. Effects of eutrophication on saltmarsh root and rhizome

biomass accumulation. Mar. Ecol. Prog. Ser. 363, 63–70 (2008).

17. Mozdzer, T. et al. Nitrogen uptake by the shoots of smooth cordgrass Spartina

alterniflora. Mar. Ecol. Prog. Ser. 433, 43–52 (2011).

18. Terzaghi, K., Peck, R. B. & Mesri, G. Soil Mechanics in Engineering Practice 3rd edn

(Wiley-Interscience, 1996).

19. Rinaldi, M., Casagli, N., Dapporto, S.& Gargini, A.Monitoring andmodelling of pore

water pressure changes and riverbank stability during flow events. Earth Surf.

Processes Landf. 29, 237–254 (2004).

20. Fagherazzi, S. et al. Numerical models of salt marsh evolution: ecological,

geomorphic, and climatic factors. Rev. Geophys. 50, RG1002 (2012).

21. Valiela, I., Teal, J. M. & Persson, N. Y. Production and dynamics of experimentally

enriched salt marsh vegetation: belowground biomass. Limnol. Oceanogr. 21,

245–252 (1976).

22. Wigand, C., Brennan, P., Stolt, M., Holt, M.&Ryba, S. Soil respiration rates in coastal

marshes subject to increased watershed nitrogen loads in southern New England,

USA. Wetlands 29, 952–963 (2009).

23. Turner, R. E. et al. Salt marshes and eutrophication: an unsustainable outcome.

Limnol. Oceanogr. 54, 1634–1642 (2009).

24. Davey, E. et al. Use of computed tomography imaging for quantifying coarse roots,

rhizomes, peat, and particle densities in marsh soils. Ecol. Appl. 21, 2156–2171


25. Morris, J. T. & Bradley, P. M. Effects of nutrient loading on the carbon balance of

coastal wetland sediments. Limnol. Oceanogr. 44, 699–702 (1999).

26. Turner, R. E. Beneath the salt marsh canopy: loss of soil strength with increasing

nutrient loads. Estuaries Coasts 34, 1084–1093 (2011).

27. Selman, M., Greenhalgh, S., Diaz, R. & Sugg, Z. Eutrophication and hypoxia in

coastal areas: a global assessment of the state of knowledge.

publication/eutrophication-and-hypoxia-in-coastal-areas (World Resources

Institute, 2008).

28. Schmidt, M. W. et al. Persistence of soil organic matter as an ecosystem property.

Nature 478, 49–56 (2011).

29. Piehler, M. F.&Smyth, A. R.Habitat-specific distinctions in estuarine denitrification

affect both ecosystem function and services. Ecosphere 2, art12 (2011).

30. Deegan, L. A. Lessons learned: the effects of nutrient enrichment on the support of

nekton by seagrass and saltmarsh ecosystems. Estuaries 25, 727–742 (2002).

31. Minello, T. J., Able, K. W., Weinstein, M. P.&Hays, C. G. Saltmarshes as nurseries for

nekton: testinghypotheses on density, growth and survival through meta-analysis.

Mar. Ecol. Prog. Ser. 246, 39–59 (2003).

32. Mariotti, G. et al. Influence of storm surges and sea level on shallow tidal basin

erosive processes. J. Geophys. Res. 115, C11012 (2010).

Supplementary Information is available in the online version of the paper.

Acknowledgements We thank our TIDE (Trophic cascades and Interacting control

processes in a Detritus-based Ecosystem), LENS (Landscape Evolution in a Nutrient

enriched Saltmarsh) and PIE-LTER (Plum Island Ecosystems Long-term Ecological

Research) colleagues for field assistance and comments. We thank the many research

assistants, graduate and undergraduate students who maintained the nutrient

enrichment and analysed samples. This work is supported by grants from the NSF

(DEB0816963, DEB0213767, OCE0923689, OCE 0423565, OCE0924287), the

NOAA and The Mellon Foundation.

Author Contributions L.A.D., D.S.J., R.S.W., B.J.P. & J.W.F. designed the experiment and

participated in sampling and data analysis. S.F. participated in geomorphic and

geotechnical evaluation. W.M.W. estimated global N loading to coastal saltmarshes.

L.A.D. and D.S.J. wrote the initial manuscript. All authors contributed to and approved

the manuscript.

Author Information The data reported in this paper are archived in the Plum Island

PIE-LTER database. Reprints and permissions information is available at The authors declare no competing financial interests.

Readers are welcome to comment on the online version of the paper. Correspondence

and requests for materials should be addressed to L.A.D. (


392 | NATURE | VOL 490 | 18 OCTOBER 2012

©2012 Macmillan Publishers Limited. All rights reserved


Site description. Our study was conducted in primary tidal creeks12,13,33,34 in the

Plum Island Estuary in northeastern Massachusetts (42u 459 N; 70u529 W). The

Plum Island Estuary is a salt-marsh-dominated systemthat is currently unaffected

by nutrient enrichment13. The system has twice-daily tides (mean tide range 2.9 m;

20–33 p.s.u.). Of the total estuarine area of 59.8km2, approximately 39.8km2

is vegetated wetlands, most of which is classic Spartina salt marsh35,36. Spartina

alterniflora (tall-form smooth cordgrass: 130–200cm in height, approximately

1,100 gm22 yr21 above-ground production) forms a twice-daily flooded 2–3-mwide

low marsh band along tidal creek channels. Spartina patens (saltmeadow

cordgrass: 20–50cm in height, approximately 600 gm22 yr21 above-ground

production) dominates the high marsh platform and is flooded by around 25%

of high tides. On creek banks and bayfronts S. alterniflora occupies the top half of

the mean tide range. S. alterniflora stems trap inorganic sediment, building steep

creek channel banks of cohesive sediment (around 15% organic matter, with the

mineral component composed of 58% silt, 25% clay and 16% sand), while belowground

roots and rhizomes (80% of root biomass in the top 20 cm, with some

rhizomes over 1m in depth) provide fibrous material that binds sediment and

stabilizes the marsh edge37–39. The percentage of organic C in the sediments of our

creek-bank marshes (7% organic C; 15% organic matter) is in the middle of

the range for salt marshes along the Atlantic and Gulf coasts of North America

(,1–28% organic C, with most marshes in the 6–15% range)40. S. patens contributes

to marsh elevation principally by below-ground biomass (80% of roots and

rhizomes are in the top 10 cm) accumulation in the form of peat.

Nutrient-enrichment experiment. Our experiment differs from others in four

important ways. (1) Nitrogen was added as nitrate (NO3

2), the form that dominates

land-derived N, contributing to coastal eutrophication and that is used

directly as an electron acceptor in decomposition processes. (2) Nutrients were

added directly to flooding tidal water to mimic the most important way in which

anthropogenic nutrient loading is delivered to tidal marshes. Dissolved nutrients

carried in water will interact and reach parts of the ecosystem differently fromdry

fertilizer. Previous plot-level salt-marsh nutrient-enrichment studies41 used dry

fertilizer (typically urea or NH4NO3) sprinkled approximately bi-weekly to

monthly on the surface of small plots (typically ,5m2) and were generally not

conducted in tall, creek-bank S. alterniflora environments. Tidal water is the primary

vector forNdelivery to coastal marshes, suggesting that if mode (dissolved in

water versus dry surface application) and type of N (NO3

2versus NH4

1) are

important in determining ecosystem response to nutrient enrichment, previous

experiments may not be sufficient for determining how salt marshes respond to

coastal nutrient enrichment. (3) We conducted a long-term (nutrient enrichment

began in 2004) ecosystem manipulation experiment in which marsh landscapes

were nutrient-enriched to levels corresponding to moderate to highly eutrophic

coastal waters42–44. Our NO3

2 enrichment target of 70–100 mMNO3

2 (added as

NaNO3) was 15 times the Plum Island background (,5 mMNO3

2) and the PO4


(added as NaH2PO4) target of 5–7 mM was 5 times the background (,1 mM


32). This was approximately a 15:1 N:P ratio in flooding creek water. (4)

We conducted an ecosystem-level experiment consisting of experimental marsh

units (n56; 3 reference and 3 nutrient-enriched) comprised of first-order creeks

(about 300mlong and 15mwide at the mouth, tapering to 2mnear terminus) and

about 30,000m2 of cordgrass marsh area, thus allowing us to examine interacting

habitats in the marsh ecosystem (creek channels, mudflats, creek-bank low marsh,

and high marsh) and examine the response of plants, animals, biogeochemical

processes and landscape-level geomorphic processes. Other factors did not

differ among treatment creeks (Supplementary Information). Analysis of baseline

characteristics before experimentalmanipulation (1998–2003) found little difference

among the experimental marsh systems13. The primary comparisons are among

two long-term (7 years by the end of 2010) nutrient-enriched (N1, N2) marshes

and two reference (R1, R2) marshes that have been intensively monitored.

To provide a time series of geomorphic change, we include data from a third

nutrient-enrichment marsh (N3) that was started in 2009.

Measurements. This is a multi-year nutrient enrichment experiment, and not all

response variables reported here were measured in each year. Some responses to

enrichment were unanticipated (for example, rapid geomorphic changes) and so

measurements were not taken before manipulation and sometimes only a single

season of data is available. The nature of the different data sets with different time

series of collections necessitated various statistical analyses (detailed below).

Above-ground plant responses. Above-ground measurements of creek-bank

S. alterniflora were taken in creeks R1, R2 and N1, N2 in years 1–7 of nutrient

enrichment. Individual plant shoot length (cm), shoot mass (g dry weight) and

shoot-specific mass (g dry weight cm21) of creek-bank S. alterniflora were measured

at the end of the growing season (mid-August) on individual shoots (18–25)

at three sites within each creek (n554–75 shoots per treatment creek per year).

Each shoot was individually washed to remove sediments, measured for length,

dried at 80 uC to a constant mass and weighed. Leaf tissue from 3–5 leaves from

each site was ground and analysed for percentage nitrogen using a PerkinElmer

2400 Series II CHNS/O analyser (n54–8 per creek per year).

In year 5 of nutrient enrichment, lignin (as a percentage of the ash-free organic

content) was determined on composite samples of 3–5 shoots from three sites

within creeks R1 and N1 as acid-insoluble fractions using a two-stage digestion in

sulphuric acid45.

Plant lodging was surveyed at the end of the season after peak production in N1,

N2, R1, R2 and four additional reference creeks in nutrient treatment year 4 (N52

for nutrient enrichment and N56 for reference). Surveys were completed on

10-m sections every 50m from the 0-m mark to 300m landward. Each section

classified into a lodging class (0–5, 5–15, 25–50 and .50% of plants in the area

lodged) for a sampling effort of n516–26 sections per creek.

Below-ground responses. Below-ground biomass, organic matter and water

content were determined by coring (n510 per creek, 10cm diameter, taken to

a depth of about 0.5 m; creeks R1, R2, N1, N2) in treatment year 7 (2010). Cores

were sliced into sections (0–5, 5–10, 10–20, 20–30, 30–40 cm), sub-samples were

taken for determination of percentage water (a small syringe core in each section),

and the remaining material was separated by sieving into two size classes of dead

organic matter (large .3mm; fine ,3mm and .1mm) and live below-ground

biomass (roots and rhizomes). Sediment geotechnical properties (percentage

water was determined as mass loss after drying a known volume of sediment at

105 uC for 24 h; percentage fine organic matter was taken to be detritus greater

than 1mmbut less than 3mmin size) were determined on cores taken for belowground

plant biomass. For statistical analysis (see ‘Statistical Summary’ below), we

focused on the top 20 cm of the cores.

Microbial decomposition processes. Total microbial production in surficial sediments

was 54% higher46 (years 1 and 2) and potential denitrification on the high

S. patens marsh was higher47 (year 3) than in the reference systems. To determine

whether microbial denitrification was also increased in creek banks, potential

denitrification48 was measured on sediment slurries—at the surface (0–5 cm)

and deep (5–10 cm)—from creek-bank cores from three sites in creeks R1 and

N1 in year 5 of nutrient enrichment (n53 per depth per creek).

To determine whether plant litter decomposition was accelerated, in year 7

plant litter respirationwasmeasured.Respirationwasmeasured fromdecomposing

litter from litterbags (15 g dry weight of S. patens; 1mmmesh size) placed flush on

the high-marsh (S. patens) surface in nutrient enrichment (N1, N2) and reference

(R1, R2) creeks. Nutrient enrichment stimulated detritivore snail densities34 and

therefore to account for the effect of detritivore density on decomposition, litterbags

were manipulated to have snail densities of 0, 1, 2, 4 or 8 times the reference

creek densities (n55–8 litterbags per creek). After 5 weeks, microbial respiration

(CO2 g21 s21) of 2–3 g of litter from the litterbags was measured using a LI-6200

Portable Photosynthesis System.

Creek-bank fracture density and vegetation loss. These measurements were

taken during the growing season in creeks R1, R2 and N1, N2 in years 6 and 7

of nutrient enrichment. Fractures, defined as a visible break in the high marsh

(S. patens-dominated) turf that parallels the creek channel (Supplementary Fig. 1)

within 3m of the S. alterniflora/S. patens border were measured early in the

growing season before cordgrass growth obscured these features. Both sides of

each creek were sampled for fractures in contiguous 50-msegments from the 0-m

mark to 200-m landward. The number of fractures and their characteristics

(length, width and depth of fracture) were recorded within each segment. In these

same segments, percentage exposed sediment (mud) area was determined by

point-intercept transects in the middle of the growing season, when the grass

canopy was fully developed, at 1-mincrements. The soil surface 1-m perpendicular

and creekward of the S. alterniflora/S. patens border was scored as ‘vegetated’ (with

S. alterniflora culms within a 30-cm-diameter circle of the point) or as ‘bare mud’

(without S. alterniflora). The fraction of points unvegetated within each

50-m segment was considered a single observation. For fracture density, fracture

length, percentage fractured and percentage exposed mud, each 50-m segment

(4 per channel side, 2 sides) was considered an individual observation, thus providing

n58 observations per creek per year. This was also done for the third

enriched and reference creeks (R3, N3) in years 1–4 of enrichment.

Creek-bank and channel structure. High-resolution total station surveys of

reference (R1, R2) and nutrient-enriched creek channels (N1, N2) and banks were

performed in years 5 (2008) and 8 (2011) of the nutrient-enrichment experiment.

We initially measured 43 cross-sections in enriched creeks and 38 in the reference

channels, with 10–20 points per cross-section, depending on the morphological

complexity of the cross-section. Twenty-six cross-sections in the nutrientenriched

creek channels and 27 in the reference channels were reoccupied after

three years, whereas in the remaining cross-sections the control poles were lost.

Erosion was computed for each cross-section as the areal difference between the

year 5 and year 8 surveys.


©2012 Macmillan Publishers Limited. All rights reserved

In year 7 of nutrient enrichment slumped sections of creek bankwere enumerated

in creek channels (R1, R2 and N1, N2). At low tide, large and small slumped

sections of creek bank were enumerated in 50-m reaches starting from the 0-m

mark up to 200–250 m. Large sections were defined as peat blocks that were

separated away from the low marsh area by at least 0.25m horizontally and at

least 0.25mlower than themean elevation of the lowmarsh. Large peat blocks were

at least 1meither in height, length or width, with small blocks being at least 0.25m

in height. Height was defined as the distance from the bottom of the peat block to

the highest point of the block. Ninety per cent of the large peat blocks had live

S. alterniflora shoots and were 1.1m wide and 2.1m long on average. Small creekbank

sections were defined as low-marsh peat chunks that were ,1m in height,

width and length, but at least 0.25m in at least one of these dimensions. Small

slumps were generally unvegetated, had visibly eroded perimeters and were

found in the deepest part of the channel. The total count of slumps per creek

was considered a sample and thus within treatment area sampling replication

was n51 per creek.

Global nutrient loading to salt marshes. We used an existing global river network

N removal model49,50 to estimate global increase in N loading to coastal

oceans from the pre-industrial period (1800s) to the contemporary period

(2000s) compared to the locations of salt marshes5,51.

Statistical summary. Analyses were performed on data primarily from the longterm

nutrient-enrichment creeks and the paired reference creeks. Data collection

from creeks often entailed sampling several subplots within experimental creeks

(see ‘Measurements’ section above). Except where noted, data were averaged

across subplots within each creek before analysis52 and statistical analyses were

performed at the creek level (nutrient enrichment n52, reference n52). All data

were checked for assumptions of normality and homoscedasticity and transformed

tomeet assumptions53. The large spatial scale of the experiment necessitated

low replication, which can reduce statistical power, so results were considered

significant at a#0.10, as is typical in a complex large-scale ecosystem experiment

where background variability is generally high and replication low54–57. Our results,

however, are robust, because 13 out of 17 tests were significant at a#0.05 despite

low replication (Table 1). Statistical analyses were performed in R (version 2.15.0).

Repeated measures analysis of variance with between-subject factors (nutrient

enrichment) were performed for response variables for which 7 years of data were

available: plant shoot height, shoot mass, shoot specific weight (natural logX11

transformation), and percentage foliar nitrogen (arcsine-squareroot transformation).

Using ‘space for time’ substitution, linear regression models were used to

analyse data for fracture density (natural logX11 transformation), fracture

length (natural logX11 transformation), percentage of creek bank with fractures

(arcsine-squareroot transformation), and percentage non-vegetated exposed mud

area (arcsine-squareroot transformation) against the number of years the creek

had received nutrient enrichment. Data for fracture density was collected in 2009–

2012 and analysis was at the creek level, thus n524: for year 0, n53 per year (R1,

R2, R3 in 2009–2012); for years 1–4, n51 per year (N3 in 2009–2012); and for

years 6–9, n52 per year (N1, N2 in 2009–2012). Data for other variables was

collected in 2009 and 2010 and analysis was at the creek level, thus n510: for year

0, n54 (two years of data from R1, R2); for year 1, n51 (N3 in 2009); for year 2,

n51 (N3 in 2010); for year 6, n52 (N1, N2 in 2009); and for year 7, n52 (N1,

N2 in 2010).

One tailed t-tests were performed on the following response variables: percentage

water content (arcsine-squareroot transformation), live below-ground

biomass, percentage fine organic matter (arcsine-squareroot transformation),

slumps per creek (logX11 transformation), lodging (percentage of plants lodged;

arcsine-squareroot transformation), potential denitrification and percent foliar

lignin (arcsine-squareroot transformation). For lodging, t-tests were performed

on the percentage of plots within a creek that were scored with .50% plants

lodged. For potential denitrification, initial statistical analysis indicated no difference

between depths, so data from different depths were pooled in the final

analysis (n56 per creek). Because potential denitrification and percentage foliar

lignin were taken only from one creek per treatment (R1 and N1), data were

analysed with the subplots as the experimental units. These process measurements

are used as supporting evidence, but cannot be used to extrapolate the results to a

wider population of systems because there was no treatment replication at the

creek level.

Analysis of covariance was performed on plant litter respiration and channel

width/depth ratio. In these analyses the factor was nutrient level—reference

(n52) and nutrient enrichment (n52). The covariate in the analysis of plant

litter respiration was the number of snails in the litterbags. The covariate in the

analysis of the channel width/depth ratio was the distance upstream from the

beginning of the treatment area (designated as 0 m; a spatial covariate). In both

analyses of covariance, the slopes were similar, but the intercepts were different,

indicating a difference between treatments at the zero covariate level.

33. Johnson, D. S. & Fleeger, J. W. Weak response of saltmarsh infauna to ecosystemwide

nutrient enrichment and fish predator reduction: a four-year study. J. Exp.

Mar. Biol. Ecol. 373, 35–44 (2009).

34. Johnson, D. S. High marsh invertebrates are susceptible to eutrophication. Mar.

Ecol. Prog. Ser. 438, 143–152 (2011).

35. Redfield, A. C. Ontogeny of a salt marsh estuary. Science 147, 50–55 (1965).

36. Niering, W. A. & Warren, R. S. Vegetation patterns and processes in New England

salt marshes. Bioscience 30, 301–307 (1980).

37. van Eerdt, M. M. The influence of vegetation on erosion and accretion in

saltmarshes of the Oosterschelde, The Netherlands. Plant Ecology 62, 367–373


38. Osman, N. & Barakbah, S. S. Parameters to predict slope stability—soil water and

root profiles. Ecol. Eng. 28, 90–95 (2006).

39. Kirwan, M. L. & Murray, A. B. A coupled geomorphic and ecological model of tidal

marsh evolution. Proc. Natl Acad. Sci. USA 104, 6118–6122 (2007).

40. Craft, C., Reader, J., Sacco, J. N. & Broome, S. W. Twenty-five years of ecosystem

development of constructed Spartina alterniflora (Loisel) marshes. Ecol. Appl. 9,

1405–1419 (1999).

41. Mendelssohn, I. A. & Morris, J. T. in Concepts and Controversies in Tidal Marsh

Ecology (eds Weinstein, M. P. & Kreeger, D. A.) 59–80 (Kluwer Academic, 2000).

42. O’Shea, M. L. & Brosnan, T. M. Trends in indicators of eutrophication in western

Long Island sound and the Hudson-Raritan Estuary. Estuaries 23, 877–901


43. Bricker, S. et al. Effects of Nutrient Enrichment In the Nation’s Estuaries: A Decade of

Change (NOAA Coastal Ocean Program Decision Analysis Series No. 26, National

Centers for Coastal Ocean Science, 2007).

44. USEnvironmental Protection Agency (USEPA). National Coastal Condition Report II. (USEPA, 2005).

45. Effland, M. J. Modified procedure to determine acid-insoluble lignin in wood and

pulp. Tappi J. 60, 143–144 (1977).

46. Bowen, J. L., Crump, B., Deegan, L. A. & Hobbie, J. E. Increased supply of ambient

nitrogen has minimal effect on salt marsh bacterial production. Limnol. Oceanogr.

54, 713–722 (2009).

47. Koop-Jakobsen, K. & Giblin, A. E. The effect of increased nitrate loading on nitrate

reduction via denitrification andDNRA in salt marsh sediments. Limnol. Oceanogr.

55, 789–802 (2010).

48. Bernot, M. J. et al.Comparing denitrification estimates for a Texas estuary by using

acetylene inhibition and membrane inlet mass spectrometry. Appl. Environ.

Microbiol. 69, 5950–5956 (2003).

49. Wollheim, W. M. et al. Global N removal by freshwater aquatic systems using a

spatially distributed, within-basin approach. Glob. Biogeochem. Cycles 22,

GB2026 (2008).

50. Beaulieu, J. J. et al. Nitrous oxide emission fromdenitrification in stream and river

networks. Proc. Natl Acad. Sci. USA 108, 214–219 (2011).

51. Murray, B. C. et al. Green Payments for Blue Carbon. Economic Incentives for

Protecting Threatened Coastal Habitats. Nicholas Institute Report NI R 11-04

(Nicholas Institute for Environmental Policy Solutions, 2011).

52. Hurlbert, S. H. Pseudoreplication and the design of ecological field experiments.

Ecol. Monogr. 54, 187–211 (1984).

53. Zar, J. H. Biostatistical Analysis 2nd edn (Prentice-Hall, 1984).

54. Carpenter, S. R., Chisholm, S. W., Krebs, C. J., Schindler, D. W. & Wright, R. F.

Ecosystem experiments. Science 269, 324–327 (1995).

55. Schindler, D. W. Replication versus realism: The need for ecosystem-scale

experiments. Ecosystems 1, 323–334 (1998).

56. Hobbs, N. T.& Hilborn, R. Alternatives to statistical hypothesis testing in ecology: A

guide to self teaching. Ecol. Appl. 16, 5–19 (2006).

57. Oksanen, L. Logic of experiments in ecology: is pseudoreplication a pseudoissue?

Oikos 94, 27–38 (2001).


©2012 Macmillan Publishers Limited. All rights reserved

Facebookby feather