Feagin et al 2005
RESEARCH COMMUNICATIONS RESEARCH COMMUNICATIONS
359
Coastal erosion, global sea-level rise, and
the loss of sand dune plant habitats
Rusty A Feagin1, Douglas J Sherman2, and William E Grant3
Much of America’s coastline is threatened by overdevelopment and coastal erosion, driven by global sea-level
rise, a problem that is attracting the attention of researchers around the world. Although we have now
acknowledged the impending risks, little is known about the response of spatially dependent dune plant com-
munities to the loss or restriction of their habitat. In order to study this development, a spatially explicit model
of sand dune plant succession on Galveston Island, Texas, was created, using sea-level rise as the primary
mechanism causing local erosion. Simulations of sea-level rise scenarios developed by the Intergovernmental
Panel on Climate Change demonstrated that beach erosion constrained plants to a narrow area, resulting in a
breakdown of the successional process. The loss of late-succession plants along coastlines, their dependent fau-
nal species, and possible solutions are discussed. This model and example serves as a harbinger of the future for
many of the US's sandy beaches and coastal communities.
Front Ecol Environ 2005; 3(7): 359–364
C oastal erosion is a problem common to shorelines
around the world, impacting about 70% of the Earth’s
sandy beach environments (Bird 1985). The causes of ero-
erosion rates, as it has historically attained the relative sea-
level rise rates that are expected to affect much of the
receding coastlines of the US over the next century.
sion can be of local (eg a decrease in sediment supply) or The Gulf Coast of the city of Galveston is protected
global importance (eg a worldwide change in sea level). from storm surge by a seawall (Figure 1a) constructed in
Beach erosion in a sediment-rich environment will result response to the damage caused by a 1900 hurricane that
in a landward displacement of coastal environments still stands as the US’s worst natural disaster (Pilkey and
(including dunes; Martinez and Psuty 2004) as long as Frasier 2003). Although the seawall and the groins that
there are no barriers (eg cliffs) to restrict such migration. front it protect the properties immediately landward,
Where there are barriers, including those commonly intro- these structures have caused greater down-drift erosion by
duced by human activities (Nordstrom 2004), erosion disrupting the natural sediment transport system, result-
reduces the area available for plant and animal communi- ing in the need for other shoreline protection measures
ties (Feagin 2005). In some places, coastal development such as geotextile tubes (Figure 1b). Along most of
confines the natural community to such a narrow stretch of Galveston Island, extensive development and the pres-
beach that plants are no longer able to disperse or grow. ence of non-native lawn vegetation behind the beach
In order to study this development, a spatially explicit presents a landward barrier that blocks the migration of
model of species dynamics within a sand dune plant com- native habitat (Figure 1c).
munity on Galveston Island, Texas, was created. Average Although it is recognized that continued erosion restricts
rates of shoreline loss on Galveston Island exceed several the plant community’s habitat within a narrow zone, it is
meters per year. Local processes such as subsidence and not yet understood what the ecological ramifications to the
reduced along-shore sediment transport have contributed plant community may be in terms of successional dynamics
to historic coastal retreat in this region (Morton et al. among species and functional groups. Moreover, there has
2004), although current subsidence rates are minimal been little research on the impact of a narrowing habitat
(USGS 2002). Erosion rates are already accelerating as a upon community pattern formation in plant succession in
result of sea-level rise (Holgate and Woodworth 2004) general. Because of the impending habitat loss of dune
induced by atmospheric warming (Titus et al. 1991; Twilley plant communities in this area, this issue is ideally suited to
et al. 2001), a problem of global dimensions (IPCC 2001). investigation through simulation modeling.
Galveston Island is an ideal site for the study of accelerated
1
Methods
Spatial Sciences Laboratory, Texas Agricultural Experiment Station
and Department of Forest Science, Texas A&M University, College Study area
Station, TX 77843-2120 (feaginr@tamu.edu); 2Department of A detailed description of the study area and baseline model,
Geography, Texas A&M University, College Station, TX 77843- minus the sea-level parameters, can be found in Feagin et al.
3147; 3Department of Wildlife and Fisheries Sciences, Texas A&M (2005). The sand dune plant community, upon which the
University, College Station, TX 77843-2258 model was based, was contained in a research plot (Figure
© The Ecological Society of America www.frontiersinecology.org
Loss of dune plant habitats RA Feagin et al.
360 (a) (b)
(c) (d)
Figure 1. Four views of the rapidly eroding sand dune plant habitat on Galveston Island, Texas. (a) Transitory dunes in front of the
seawall, just visible beyond the lip of the wall; (b) habitat in front of geotextile tubes during a natural storm event (here, Hurricane
Lily, the eye of which passed over Louisiana, 400 km to the east); and (c) the maintenance of non-native lawn vegetation, leaving a
tenuous 2 m strip of habitat. (d) Under natural circumstances, the native plant community in the research plot can “move” with the
dunes and the back-and-forth migration of the beach.
1d) adjacent to Pirates Beach on Galveston Island, at lati- mented in Mathematica (Wolfram 1996). A two-dimen-
tude 29.21 and longitude –94.92. The climate at this loca- sional lattice, composed of 100 x 100 individual sites, was
tion is subtropical, characterized by hot summers and mild created. Each site represented a 0.5 x 0.5 m (0.25 m2)
winters. The community extends along a 50 m beach–sand quadrat in the research plot. The top of the lattice repre-
dune gradient, stretching from the summer berm on the sented the landward boundary of the sand dune plant
open beach to the edge of the upland/coastal prairie com- community while the bottom of the lattice represented
munity behind the dune ridge. The sand dune plant com- the seaward boundary.
munity, depicted in Figure 1d, can be described in terms of Each site had four possible states, consisting of bare-
its three successional stages: colonizers, soil binders, and ground (0), colonizer plants (1), soil binder plants (2), or
competitors. Multiple years of plant succession in this com- competitive plants (3). Because sand dune succession is
munity have been documented since 1998, when Tropical based on facilitation by adjacent plants, as well as the
Storm Francis destroyed the plant cover and subsurface amelioration of environmental constraints by previous
growth, removing the soil and leaving only bare sand. The successional stages, each state represented one step in the
model was calibrated based on spatial patterns observed successional sequence. For example, a site could start as
over a 5-year period at this location, with the beach accret- bareground and move through the three subsequent suc-
ing at an average rate and growing conditions that were typ- cessional stages if an adequate number of other plants
ical for the subtropical, humid coastal dunes of Texas. were growing in adjacent locations. The relative effect of
a site upon an adjacent site, through facilitation, was
Baseline model description based on proximity, as detailed in Feagin et al. (2005).
During each time step, all sites were simultaneously
The sand dune plant community was modeled using two- updated based on the adjacent sites, with time steps rep-
dimensional cellular automata, and written and imple- resenting 6 months. The model was run for a total of
www.frontiersinecology.org © The Ecological Society of America
RA Feagin et al. Loss of dune plant habitats
6 years, as this is the frequency of catastrophic distur- upon coastal erosion rates at specific locations. 361
bances expected at the research site (Morton et al. 1983). However, the Bruun Rule could not be used indepen-
Such disturbances often result in the retrogressing of the dently, as it does not accommodate alongshore variability
site back to its initial bareground state. in profile response (eg Stive 2004). To address this vari-
An environmental stress gradient was also constructed, ability, we used historical erosion rate data, available in
where stress values (–1) were distributed along the gradi- a spatially explicit layer at 50 m increments in the
ent, with 100% of the sites undergoing stress at the sea- longshore direction (Gibeaut et al. 2003), and the histor-
ward edge and 0% at the landward edge. Sites could ret- ical eustatic rise rate from the tidal gauge data, and
rogress to earlier stages of succession if the summation of applied the Bruun Rule every 50 m along the beach in
adjacent sites was reduced by adding stress. The baseline order to derive the profile response ratio. We then
model produced plant patterns which approximated the assumed that the ratio represented a constant function
research plot data, as described in Feagin et al. (2005). describing the cross-shore equilibrium profile at each
location, and back-calculated the expected erosion rate
Sea-level rise calculations for the projected sea-level rise scenarios.
For example, the historical data yields:
One estimate for the global rate of sea-level rise is around
0.20 cm yr–1 over the past 100 years. However, the histori- s 1.68 m yr–1
cal local rate may differ greatly as the relative rise is a func- = = 509
a 0.0033m yr–1
tion of several components, including eustatic (eg global
sea-level rise due to deglaciation, or thermal expansion of
water at higher temperatures), isostatic (eg land elevation as the ratio at our research plot, the value that we then
changes caused by plate tectonics, or rebound of land after used for simulations at the plot site. If we assume eustatic
deglaciation), and subsidence effects (caused by with- sea-level rises at 1.00 cm yr–1, then we calculate:
drawal of water, oil, and gas, or compaction of sediments;
Douglas 1991). The local rate of relative rise at the s x
research plot was 0.65 cm yr–1 between 1909 and 1999, as = = 509
a 0.01m yr–1
recorded by the Galveston tide gauge (Permanent Service
for Mean Sea Level; www.pole.ac.uk/psmsl). During most
of this period, the area experienced dramatic land subsi- with the projected erosion rate at that spatial location as
dence, at a rate of 0.32 cm yr–1, due to oil and groundwater x = 5.09 m yr–1. We repeated this calculation every 50 m
extraction (USGS 2002; www.hg.subsidence.org). If we in the longshore direction, using the local values for the
subtract 0.32 from 0.65, this yields 0.33 cm yr–1 as the rate Bruun ratio.
of eustatic rise at the research plot. This number exceeds This method assumes that the rate and direction of long-
the global eustatic rate, but there are two possible explana- shore sediment transport does not change when the sea-
tions for this: (1) thermal expansion in the warm Gulf of level rises, yet it accounts for morphologically different
Mexico (Miller and Douglas 2004), which agrees quite cross-shore profiles at spatial locations in the longshore
well with Leatherman’s (1984) calculations for this area, direction. Our method assumes sediment transport in the
and/or (2) a slight, natural isostatic change at a much erosion term, s, by using the Gibeaut et al. (2003) dataset at
larger regional scale than that of the USGS (2002) data. each longshore location. As a reasonable first approxima-
Sea-level rise is a dominant force driving widespread tion in our model, the proportion of s that could be due to
coastal erosion (Leatherman et al. 2000; Zhang et al. 2004). a negative or positive sediment transport term (often
Since the terrain of the Gulf Coast is quite flat, a small rise known as q in coastal engineering literature) will propor-
in water level can result in a severe loss of land. Historical tionately change with the rise, a, based on the empirically
erosion rates exceed 3.3 m yr–1 at the research plot loca- derived relationship at each spatial location separately.
tion (Gibeaut et al. 2003; www.beg.utexas.edu/coastal/haz- Another assumption is that sediment particle sizes remain
ard_atlas1.htm). If we partition this erosion rate into a lin- constant, also subsumed in the relatively high Bruun rule
ear function of its known subsidence and eustatic ratios that can be found on Galveston’s high clay content
components, then 1.68 m yr–1 of this erosion was due to beaches. The method also assumes that the cause of histor-
eustatic rise. ical, long-term erosion is sea-level rise, with beach profiles
To simulate the impacts of sea-level rise on coastal recovering from transitory storm effects; this accords with
retreat, we extrapolated the historical erosion rate into the most recent evidence (Zhang et al. 2002).
the future, and assumed that the Bruun Rule (essentially
a ratio of the sea level rise rate, a, to the profile migration, Simulations
or erosion, rate, s) provides a reasonable first approxima-
tion of cross-shore profile response of water-level change Simulations expanded upon the baseline model by (1) test-
(Leatherman et al. 2000; Zhang et al. 2004). We then pro- ing hypotheses about plant community response to erosion
jected future sea-level rise scenarios and their impacts induced by climate change, and (2) adding new parameters
© The Ecological Society of America www.frontiersinecology.org
Loss of dune plant habitats RA Feagin et al.
362 (a) Several changes were also made to facilitate
investigations into plant community responses
to longshore morphology. First, the scale of
inquiry was broadened and 1500 x 100 m strips
were demarcated along the coastline on an aer-
ial photograph within a geographic information
system (GIS). A grid was created, showing
locations where dune growth was restricted by
Relative sea-level barriers such seawalls, geotextile tubes, mani-
rise risk cured lawn vegetation, or housing develop-
high ments. The grid was then fed into the model
and run as a 3000 x 200 site lattice, with each
low site representing 0.5 x 0.5 m in the GIS. The
(b) historical erosion rate data were also fed into
the model with values explicitly known at 50 m
increments in the longshore direction. In addi-
tion to the previous four states (0, 1, 2, 3), sites
could now also be categorized as an upland
plant community (4) or a barrier to dune devel-
opment (5). Model output was then geo-refer-
enced within the GIS, producing maps that
detailed plant community responses to the
(c) (d) (e) much broader-scaled phenomena of longshore
coastal morphology.
Results and discussion
Coastal erosion due to sea-level rise, in concert
with human-erected barriers, confined simu-
lated plants to a small zone at the rear of the
Sand Soil binders Upland/coastal prairie
research plot, altering the sand dune plant
Colonizers Competitors Barrier
community’s characteristic patterns. Although
littoral drift, sediment dynamics, and beach
Figure 2. Sea-level rise, the risks, and impacts to the sand dune plant composition may differentially affect the risks
community at multiple scales. (a) The risk of relative sea-level rise at the scale to sand dune habitat at a continental scale
of the continental US (adapted source data from Thieler and Hammar-Klose (Figure 2a), the landscape-scale GIS output
2000). At the Galveston Island landscape scale (b), the model was run and demonstrates little deviation in the plant com-
geo-referenced within a GIS. Here, we show the results for the most munity’s development as a result of the long-
accelerated rate of sea-level rise, with the circles representing the spatial shore morphology response to sea-level rise, for
locations mentioned in the text. At the plant community scale, simulations 50 m increments (Figure 2b). Yet the spatial
were executed under the three IPCC scenarios: the (c) low, (d) moderate, and limitations enforced by the barriers led to a dif-
(e) high rise scenarios. Plant community scale maps have the ocean at the ferential placement of the community in the
bottom, a barrier at the top, and are 35 x 35 m. longshore direction, leaving some areas com-
pletely barren of dune habitat and others with
to the model to incorporate longshore morphological char- remnant dune populations.
acteristics at a broader scale with the help of a Geographic Strong differences in community pattern development
Information System (GIS). Simulations tested the effect of resulted in the cross-shore direction under the three IPCC
sea-level rise on the plant patterns by moving the stress scenarios. The low rise scenario (Figure 2c) allowed the
gradient in the landward direction along the cross-shore plant community to develop fully in 5 years within the
profile. The rate at which the rising sea, and the associated research plot area. Dunes were covered by all stages of suc-
stress gradient, moved landward was represented in three cession, including coastal prairie and upland plants.
well-known Intergovernmental Panel on Climate Change Under the moderate rise scenario (Figure 2d), plants did
simulation scenarios (IPCC 2001): a low rise scenario not grow in the lower section of the gradient due to high
(0.09 m by 2100), a moderate rise scenario (0.48 m by stress levels. There was a random distribution of colonizers
2100), and a high rise scenario (0.88 m by 2100). We mod- across the mid-section of the gradient, yet the community
eled each scenario according to the method outlined in the did not develop further due to the high level of stress and
previous section. the lack of permanence in the distribution of the annual
www.frontiersinecology.org © The Ecological Society of America
RA Feagin et al. Loss of dune plant habitats
colonizers. In the upper section of the gradient, only a few certain to increase. Dependent animal species, such as 363
embryonic dunes had formed, around which the later suc- the endangered Kemp’s ridley sea turtle (Lepidochelys
cessional species could coalesce. In these areas, the plant kempii) and the endangered Choctawatchee beach mouse
distribution was not random, as each successional stage (Peromyscus polionotus), will be lost without them.
depended on prior amelioration of the location by the col- Although our simulations are restricted to the conditions
onizers in order to survive. at Galveston Island, the basis for the model suggests that
The high rise scenario resulted in a complete breakdown similar scenarios, with different retreat rates, will occur in
of the successional process, as only colonizers were ran- the near future, in most dune systems along developed
domly distributed and located in the upper section of the coastlines experiencing sea-level transgression.
gradient in the research plot area (Figure 2e). The random Considerable planning needs to be done in order to
distribution of the annual colonizers, their ephemeral decide how we want our coasts to look, as well as to pre-
nature, and their restriction to the thin strip of habitat did serve the species that inhabit them. A small increase in
not provide the ameliorative force (eg windblocks, islands sea-level rise can result in a large amount of coastal ero-
of fertility, elevated dune structures) that was needed for sion. It may soon be necessary to reassess the importance
the community to progress to advanced stages of succes- of sand dune plant communities, in the same way that
sion. In terms of the spatial and temporal mechanics of coastal wetlands were re-evaluated by the last generation
any one focal site, the summation of its adjacent positively of coastal zone statutes and laws. As it is our duty to
valued sites (facilitative plants) and its adjacent nega- maintain these coastal plant communities alongside our
tively valued sites (stress) produced a number that did not private developments and shoreline protection structures,
exceed the threshold for its graduation to the next stage of we must publicly consider the impacts of squeezing the
succession. Without strong and spatially explicit facilita- sand dune plant habitat between the land and the sea.
tion to counterbalance the stress induced by sea-level rise,
each adjacent site had no lasting temporal effect upon this References
focal site. The successional process became decoupled; the Bird ECF. 1985. Coastline changes: a global review. Chichester,
dynamics within the sites at the upper part of the gradient UK: Wiley and Sons.
became spatially isolated and temporally ephemeral. In Douglas BC. 1991. Global sea level rise. J Geophys Res 96:
6981–92.
the lower part of the gradient, it was impossible for even Feagin RA. 2005. Artificial dunes created to protect property on
the colonizers to survive the stress. Galveston Island, Texas: the lessons learned. Ecol Restor 23:
The model produced graphical output which matched 89–94.
quite well with what has already been seen along the Gulf Feagin RA, Wu XB, Smeins FE, et al. 2005. Individual versus
Coast (Figures 1a–c) and the East Coast of the US community-level processes and pattern formation in a model
of sand dune plant succession. Ecol Model 183: 435–49.
(Roman and Nordstrom 1988); late-succession sand dune Gibeaut JC, Hepner TL, Waldinger R, et al. 2003. Geotubes for
species are being lost because embryonic dunes are not temporary erosion control and storm surge protection along
able to form in front of the barriers (Gibeaut et al. 2003), the Gulf of Mexico shoreline of Texas. Proceedings of the
leaving isolated “islands” of plant communities (Figure 13th Biennial Coastal Zone Conference; 13–17 July 2003;
2b) that are spatially removed from other communities, Baltimore, Maryland: NOAA. www.beg.utexas.edu/coastal/
hazard_atlas1.htm. Viewed 28 June 2005.
sometimes by considerable distances. Greipsson S. 2002. Coastal dunes. In: Perrow MR and Davy AJ
In the short term, money and time must be spent to (Eds). Handbook of ecological restoration. Volume 2:
restore these communities. In many cases, late-succession Restoration in practice. Cambridge, UK: Cambridge
species such as beach panic (Panicum amarum) have very University Press.
low rates of seed viability, and must therefore be planted Holgate SJ and Woodworth PL. 2004. Evidence for enhanced
coastal sea level rise during the 1990s. Geophys Res Lett 31:
during restoration efforts in order for them to appear L07305.
(Feagin 2005). The loss of such species is already occur- IPCC (Intergovernmental Panel on Climate Change). 2001.
ring on Galveston Island, where sea oats (Uniola panicu- Climate change 2001: impacts, adaptation, and vulnerability.
lata) has disappeared due to a combination of human- In: McCarthy JJ, Canziani OF, Leary NA, et al. (Eds).
induced disturbance and climate change (Greipsson Contribution of Working Group II to the Third Assesment
Report of the Intergovernmental Panel on Climate Change.
2002). In the long term, it may be critical to maintain New York, NY: Cambridge University Press.
propagule dispersal sites with planning done at a land- Leatherman SP. 1984. Coastal geomorphic responses to sea-level
scape scale in order to maintain populations along the rise: Galveston Bay, Texas. In: Barth MC and Titus JG (Eds).
East Coast of the US and in areas of large-scale develop- Greenhouse effect and sea-level rise. New York, NY: Van
ment, even for early succession species such as the endan- Nostrand Reinhold.
Leatherman SP, Zhang K, and Douglas BC. 2000. Sea level rise
gered sea-beach amaranth (Amaranthus pumilus). shown to drive coastal erosion. Eos 81: 55–57.
The maintenance of the late-succession species is criti- Martinez ML and Psuty NP. 2004. Coastal dunes: ecology and
cal, as they are usually the most important species in the conservation. New York, NY: Springer.
building of dunes, binding of sediments, and reduction of Miller L and Douglas BC. 2004. Mass and volume contributions
erosion. Moreover, these perennial species provide cover to twentieth-century global sea level rise. Nature 428:
406–09.
on the dunes throughout the year; if lost, erosion rates are
© The Ecological Society of America www.frontiersinecology.org
Loss of dune plant habitats RA Feagin et al.
Morton RA, Miller TL, and Moore LJ. 2004. National assessment Report 00–179. http://pubs.usgs.gov/dds/dds68/. Viewed 28
364 of shoreline change: Part 1, historical shoreline changes and June 2005.
associated coastal land loss along the US Gulf of Mexico. Titus JG, Park RA, Leatherman SP, et al. 1991. Greenhouse effect
USGS Open File Report 2004–1043. and sea level rise: the cost of holding back the sea. Coast
Morton RA, Pilkey Jr OH, Pilkey Sr OH, and Neal WJ. 1983. Manage 19: 171–204.
Living with the Texas shore. Durham, NC: Duke University Twilley RR, Barron EJ, Gholz HL, et al. 2001. Confronting climate
Press. change in the Gulf Coast region: prospects for sustaining our eco-
Nordstrom KF. 2004. Beaches and dunes of developed coasts. logical heritage. Cambridge, MA, and Washington, DC: Union
Cambridge, UK: Cambridge University Press. of Concerned Scientists and Ecological Society of America.
Pilkey OH and Frasier ME. 2003. A celebration of the world’s USGS. 2002. Houston–Galveston Bay area, Texas, from space – a
barrier islands. New York, NY: Columbia University Press. new tool for mapping land subsidence. Washington, DC:
Roman CT and Nordstrom KF. 1988. The effect of erosion rate on USGS. Fact Sheet 110–02.
vegetation patterns of an East Coast Barrier Island. Estuar Wolfram S. 1996. The Mathematica book. Cambridge, MA:
Coast Shelf S 26: 233–42. Wolfram Media/Cambridge University Press.
Stive MJF. 2004. How important is global warming for coastal Zhang K, Douglas BC, and Leatherman SP. 2002. Do storms cause
erosion? Climatic Change 64: 27–39. long-term beach erosion along the US East Barrier Coast? J
Thieler ER and Hammar-Klose ES. 2000. National assessment of Geol 110: 493–502.
coastal vulnerability to future sea-level rise: preliminary Zhang K, Douglas BC, and Leatherman SP. 2004. Global warming
results for the US Gulf of Mexico coast. USGS Open-File and coastal erosion. Climatic Change 64: 41–58.
www.frontiersinecology.org © The Ecological Society of America
359
Coastal erosion, global sea-level rise, and
the loss of sand dune plant habitats
Rusty A Feagin1, Douglas J Sherman2, and William E Grant3
Much of America’s coastline is threatened by overdevelopment and coastal erosion, driven by global sea-level
rise, a problem that is attracting the attention of researchers around the world. Although we have now
acknowledged the impending risks, little is known about the response of spatially dependent dune plant com-
munities to the loss or restriction of their habitat. In order to study this development, a spatially explicit model
of sand dune plant succession on Galveston Island, Texas, was created, using sea-level rise as the primary
mechanism causing local erosion. Simulations of sea-level rise scenarios developed by the Intergovernmental
Panel on Climate Change demonstrated that beach erosion constrained plants to a narrow area, resulting in a
breakdown of the successional process. The loss of late-succession plants along coastlines, their dependent fau-
nal species, and possible solutions are discussed. This model and example serves as a harbinger of the future for
many of the US's sandy beaches and coastal communities.
Front Ecol Environ 2005; 3(7): 359–364
C oastal erosion is a problem common to shorelines
around the world, impacting about 70% of the Earth’s
sandy beach environments (Bird 1985). The causes of ero-
erosion rates, as it has historically attained the relative sea-
level rise rates that are expected to affect much of the
receding coastlines of the US over the next century.
sion can be of local (eg a decrease in sediment supply) or The Gulf Coast of the city of Galveston is protected
global importance (eg a worldwide change in sea level). from storm surge by a seawall (Figure 1a) constructed in
Beach erosion in a sediment-rich environment will result response to the damage caused by a 1900 hurricane that
in a landward displacement of coastal environments still stands as the US’s worst natural disaster (Pilkey and
(including dunes; Martinez and Psuty 2004) as long as Frasier 2003). Although the seawall and the groins that
there are no barriers (eg cliffs) to restrict such migration. front it protect the properties immediately landward,
Where there are barriers, including those commonly intro- these structures have caused greater down-drift erosion by
duced by human activities (Nordstrom 2004), erosion disrupting the natural sediment transport system, result-
reduces the area available for plant and animal communi- ing in the need for other shoreline protection measures
ties (Feagin 2005). In some places, coastal development such as geotextile tubes (Figure 1b). Along most of
confines the natural community to such a narrow stretch of Galveston Island, extensive development and the pres-
beach that plants are no longer able to disperse or grow. ence of non-native lawn vegetation behind the beach
In order to study this development, a spatially explicit presents a landward barrier that blocks the migration of
model of species dynamics within a sand dune plant com- native habitat (Figure 1c).
munity on Galveston Island, Texas, was created. Average Although it is recognized that continued erosion restricts
rates of shoreline loss on Galveston Island exceed several the plant community’s habitat within a narrow zone, it is
meters per year. Local processes such as subsidence and not yet understood what the ecological ramifications to the
reduced along-shore sediment transport have contributed plant community may be in terms of successional dynamics
to historic coastal retreat in this region (Morton et al. among species and functional groups. Moreover, there has
2004), although current subsidence rates are minimal been little research on the impact of a narrowing habitat
(USGS 2002). Erosion rates are already accelerating as a upon community pattern formation in plant succession in
result of sea-level rise (Holgate and Woodworth 2004) general. Because of the impending habitat loss of dune
induced by atmospheric warming (Titus et al. 1991; Twilley plant communities in this area, this issue is ideally suited to
et al. 2001), a problem of global dimensions (IPCC 2001). investigation through simulation modeling.
Galveston Island is an ideal site for the study of accelerated
1
Methods
Spatial Sciences Laboratory, Texas Agricultural Experiment Station
and Department of Forest Science, Texas A&M University, College Study area
Station, TX 77843-2120 (feaginr@tamu.edu); 2Department of A detailed description of the study area and baseline model,
Geography, Texas A&M University, College Station, TX 77843- minus the sea-level parameters, can be found in Feagin et al.
3147; 3Department of Wildlife and Fisheries Sciences, Texas A&M (2005). The sand dune plant community, upon which the
University, College Station, TX 77843-2258 model was based, was contained in a research plot (Figure
© The Ecological Society of America www.frontiersinecology.org
Loss of dune plant habitats RA Feagin et al.
360 (a) (b)
(c) (d)
Figure 1. Four views of the rapidly eroding sand dune plant habitat on Galveston Island, Texas. (a) Transitory dunes in front of the
seawall, just visible beyond the lip of the wall; (b) habitat in front of geotextile tubes during a natural storm event (here, Hurricane
Lily, the eye of which passed over Louisiana, 400 km to the east); and (c) the maintenance of non-native lawn vegetation, leaving a
tenuous 2 m strip of habitat. (d) Under natural circumstances, the native plant community in the research plot can “move” with the
dunes and the back-and-forth migration of the beach.
1d) adjacent to Pirates Beach on Galveston Island, at lati- mented in Mathematica (Wolfram 1996). A two-dimen-
tude 29.21 and longitude –94.92. The climate at this loca- sional lattice, composed of 100 x 100 individual sites, was
tion is subtropical, characterized by hot summers and mild created. Each site represented a 0.5 x 0.5 m (0.25 m2)
winters. The community extends along a 50 m beach–sand quadrat in the research plot. The top of the lattice repre-
dune gradient, stretching from the summer berm on the sented the landward boundary of the sand dune plant
open beach to the edge of the upland/coastal prairie com- community while the bottom of the lattice represented
munity behind the dune ridge. The sand dune plant com- the seaward boundary.
munity, depicted in Figure 1d, can be described in terms of Each site had four possible states, consisting of bare-
its three successional stages: colonizers, soil binders, and ground (0), colonizer plants (1), soil binder plants (2), or
competitors. Multiple years of plant succession in this com- competitive plants (3). Because sand dune succession is
munity have been documented since 1998, when Tropical based on facilitation by adjacent plants, as well as the
Storm Francis destroyed the plant cover and subsurface amelioration of environmental constraints by previous
growth, removing the soil and leaving only bare sand. The successional stages, each state represented one step in the
model was calibrated based on spatial patterns observed successional sequence. For example, a site could start as
over a 5-year period at this location, with the beach accret- bareground and move through the three subsequent suc-
ing at an average rate and growing conditions that were typ- cessional stages if an adequate number of other plants
ical for the subtropical, humid coastal dunes of Texas. were growing in adjacent locations. The relative effect of
a site upon an adjacent site, through facilitation, was
Baseline model description based on proximity, as detailed in Feagin et al. (2005).
During each time step, all sites were simultaneously
The sand dune plant community was modeled using two- updated based on the adjacent sites, with time steps rep-
dimensional cellular automata, and written and imple- resenting 6 months. The model was run for a total of
www.frontiersinecology.org © The Ecological Society of America
RA Feagin et al. Loss of dune plant habitats
6 years, as this is the frequency of catastrophic distur- upon coastal erosion rates at specific locations. 361
bances expected at the research site (Morton et al. 1983). However, the Bruun Rule could not be used indepen-
Such disturbances often result in the retrogressing of the dently, as it does not accommodate alongshore variability
site back to its initial bareground state. in profile response (eg Stive 2004). To address this vari-
An environmental stress gradient was also constructed, ability, we used historical erosion rate data, available in
where stress values (–1) were distributed along the gradi- a spatially explicit layer at 50 m increments in the
ent, with 100% of the sites undergoing stress at the sea- longshore direction (Gibeaut et al. 2003), and the histor-
ward edge and 0% at the landward edge. Sites could ret- ical eustatic rise rate from the tidal gauge data, and
rogress to earlier stages of succession if the summation of applied the Bruun Rule every 50 m along the beach in
adjacent sites was reduced by adding stress. The baseline order to derive the profile response ratio. We then
model produced plant patterns which approximated the assumed that the ratio represented a constant function
research plot data, as described in Feagin et al. (2005). describing the cross-shore equilibrium profile at each
location, and back-calculated the expected erosion rate
Sea-level rise calculations for the projected sea-level rise scenarios.
For example, the historical data yields:
One estimate for the global rate of sea-level rise is around
0.20 cm yr–1 over the past 100 years. However, the histori- s 1.68 m yr–1
cal local rate may differ greatly as the relative rise is a func- = = 509
a 0.0033m yr–1
tion of several components, including eustatic (eg global
sea-level rise due to deglaciation, or thermal expansion of
water at higher temperatures), isostatic (eg land elevation as the ratio at our research plot, the value that we then
changes caused by plate tectonics, or rebound of land after used for simulations at the plot site. If we assume eustatic
deglaciation), and subsidence effects (caused by with- sea-level rises at 1.00 cm yr–1, then we calculate:
drawal of water, oil, and gas, or compaction of sediments;
Douglas 1991). The local rate of relative rise at the s x
research plot was 0.65 cm yr–1 between 1909 and 1999, as = = 509
a 0.01m yr–1
recorded by the Galveston tide gauge (Permanent Service
for Mean Sea Level; www.pole.ac.uk/psmsl). During most
of this period, the area experienced dramatic land subsi- with the projected erosion rate at that spatial location as
dence, at a rate of 0.32 cm yr–1, due to oil and groundwater x = 5.09 m yr–1. We repeated this calculation every 50 m
extraction (USGS 2002; www.hg.subsidence.org). If we in the longshore direction, using the local values for the
subtract 0.32 from 0.65, this yields 0.33 cm yr–1 as the rate Bruun ratio.
of eustatic rise at the research plot. This number exceeds This method assumes that the rate and direction of long-
the global eustatic rate, but there are two possible explana- shore sediment transport does not change when the sea-
tions for this: (1) thermal expansion in the warm Gulf of level rises, yet it accounts for morphologically different
Mexico (Miller and Douglas 2004), which agrees quite cross-shore profiles at spatial locations in the longshore
well with Leatherman’s (1984) calculations for this area, direction. Our method assumes sediment transport in the
and/or (2) a slight, natural isostatic change at a much erosion term, s, by using the Gibeaut et al. (2003) dataset at
larger regional scale than that of the USGS (2002) data. each longshore location. As a reasonable first approxima-
Sea-level rise is a dominant force driving widespread tion in our model, the proportion of s that could be due to
coastal erosion (Leatherman et al. 2000; Zhang et al. 2004). a negative or positive sediment transport term (often
Since the terrain of the Gulf Coast is quite flat, a small rise known as q in coastal engineering literature) will propor-
in water level can result in a severe loss of land. Historical tionately change with the rise, a, based on the empirically
erosion rates exceed 3.3 m yr–1 at the research plot loca- derived relationship at each spatial location separately.
tion (Gibeaut et al. 2003; www.beg.utexas.edu/coastal/haz- Another assumption is that sediment particle sizes remain
ard_atlas1.htm). If we partition this erosion rate into a lin- constant, also subsumed in the relatively high Bruun rule
ear function of its known subsidence and eustatic ratios that can be found on Galveston’s high clay content
components, then 1.68 m yr–1 of this erosion was due to beaches. The method also assumes that the cause of histor-
eustatic rise. ical, long-term erosion is sea-level rise, with beach profiles
To simulate the impacts of sea-level rise on coastal recovering from transitory storm effects; this accords with
retreat, we extrapolated the historical erosion rate into the most recent evidence (Zhang et al. 2002).
the future, and assumed that the Bruun Rule (essentially
a ratio of the sea level rise rate, a, to the profile migration, Simulations
or erosion, rate, s) provides a reasonable first approxima-
tion of cross-shore profile response of water-level change Simulations expanded upon the baseline model by (1) test-
(Leatherman et al. 2000; Zhang et al. 2004). We then pro- ing hypotheses about plant community response to erosion
jected future sea-level rise scenarios and their impacts induced by climate change, and (2) adding new parameters
© The Ecological Society of America www.frontiersinecology.org
Loss of dune plant habitats RA Feagin et al.
362 (a) Several changes were also made to facilitate
investigations into plant community responses
to longshore morphology. First, the scale of
inquiry was broadened and 1500 x 100 m strips
were demarcated along the coastline on an aer-
ial photograph within a geographic information
system (GIS). A grid was created, showing
locations where dune growth was restricted by
Relative sea-level barriers such seawalls, geotextile tubes, mani-
rise risk cured lawn vegetation, or housing develop-
high ments. The grid was then fed into the model
and run as a 3000 x 200 site lattice, with each
low site representing 0.5 x 0.5 m in the GIS. The
(b) historical erosion rate data were also fed into
the model with values explicitly known at 50 m
increments in the longshore direction. In addi-
tion to the previous four states (0, 1, 2, 3), sites
could now also be categorized as an upland
plant community (4) or a barrier to dune devel-
opment (5). Model output was then geo-refer-
enced within the GIS, producing maps that
detailed plant community responses to the
(c) (d) (e) much broader-scaled phenomena of longshore
coastal morphology.
Results and discussion
Coastal erosion due to sea-level rise, in concert
with human-erected barriers, confined simu-
lated plants to a small zone at the rear of the
Sand Soil binders Upland/coastal prairie
research plot, altering the sand dune plant
Colonizers Competitors Barrier
community’s characteristic patterns. Although
littoral drift, sediment dynamics, and beach
Figure 2. Sea-level rise, the risks, and impacts to the sand dune plant composition may differentially affect the risks
community at multiple scales. (a) The risk of relative sea-level rise at the scale to sand dune habitat at a continental scale
of the continental US (adapted source data from Thieler and Hammar-Klose (Figure 2a), the landscape-scale GIS output
2000). At the Galveston Island landscape scale (b), the model was run and demonstrates little deviation in the plant com-
geo-referenced within a GIS. Here, we show the results for the most munity’s development as a result of the long-
accelerated rate of sea-level rise, with the circles representing the spatial shore morphology response to sea-level rise, for
locations mentioned in the text. At the plant community scale, simulations 50 m increments (Figure 2b). Yet the spatial
were executed under the three IPCC scenarios: the (c) low, (d) moderate, and limitations enforced by the barriers led to a dif-
(e) high rise scenarios. Plant community scale maps have the ocean at the ferential placement of the community in the
bottom, a barrier at the top, and are 35 x 35 m. longshore direction, leaving some areas com-
pletely barren of dune habitat and others with
to the model to incorporate longshore morphological char- remnant dune populations.
acteristics at a broader scale with the help of a Geographic Strong differences in community pattern development
Information System (GIS). Simulations tested the effect of resulted in the cross-shore direction under the three IPCC
sea-level rise on the plant patterns by moving the stress scenarios. The low rise scenario (Figure 2c) allowed the
gradient in the landward direction along the cross-shore plant community to develop fully in 5 years within the
profile. The rate at which the rising sea, and the associated research plot area. Dunes were covered by all stages of suc-
stress gradient, moved landward was represented in three cession, including coastal prairie and upland plants.
well-known Intergovernmental Panel on Climate Change Under the moderate rise scenario (Figure 2d), plants did
simulation scenarios (IPCC 2001): a low rise scenario not grow in the lower section of the gradient due to high
(0.09 m by 2100), a moderate rise scenario (0.48 m by stress levels. There was a random distribution of colonizers
2100), and a high rise scenario (0.88 m by 2100). We mod- across the mid-section of the gradient, yet the community
eled each scenario according to the method outlined in the did not develop further due to the high level of stress and
previous section. the lack of permanence in the distribution of the annual
www.frontiersinecology.org © The Ecological Society of America
RA Feagin et al. Loss of dune plant habitats
colonizers. In the upper section of the gradient, only a few certain to increase. Dependent animal species, such as 363
embryonic dunes had formed, around which the later suc- the endangered Kemp’s ridley sea turtle (Lepidochelys
cessional species could coalesce. In these areas, the plant kempii) and the endangered Choctawatchee beach mouse
distribution was not random, as each successional stage (Peromyscus polionotus), will be lost without them.
depended on prior amelioration of the location by the col- Although our simulations are restricted to the conditions
onizers in order to survive. at Galveston Island, the basis for the model suggests that
The high rise scenario resulted in a complete breakdown similar scenarios, with different retreat rates, will occur in
of the successional process, as only colonizers were ran- the near future, in most dune systems along developed
domly distributed and located in the upper section of the coastlines experiencing sea-level transgression.
gradient in the research plot area (Figure 2e). The random Considerable planning needs to be done in order to
distribution of the annual colonizers, their ephemeral decide how we want our coasts to look, as well as to pre-
nature, and their restriction to the thin strip of habitat did serve the species that inhabit them. A small increase in
not provide the ameliorative force (eg windblocks, islands sea-level rise can result in a large amount of coastal ero-
of fertility, elevated dune structures) that was needed for sion. It may soon be necessary to reassess the importance
the community to progress to advanced stages of succes- of sand dune plant communities, in the same way that
sion. In terms of the spatial and temporal mechanics of coastal wetlands were re-evaluated by the last generation
any one focal site, the summation of its adjacent positively of coastal zone statutes and laws. As it is our duty to
valued sites (facilitative plants) and its adjacent nega- maintain these coastal plant communities alongside our
tively valued sites (stress) produced a number that did not private developments and shoreline protection structures,
exceed the threshold for its graduation to the next stage of we must publicly consider the impacts of squeezing the
succession. Without strong and spatially explicit facilita- sand dune plant habitat between the land and the sea.
tion to counterbalance the stress induced by sea-level rise,
each adjacent site had no lasting temporal effect upon this References
focal site. The successional process became decoupled; the Bird ECF. 1985. Coastline changes: a global review. Chichester,
dynamics within the sites at the upper part of the gradient UK: Wiley and Sons.
became spatially isolated and temporally ephemeral. In Douglas BC. 1991. Global sea level rise. J Geophys Res 96:
6981–92.
the lower part of the gradient, it was impossible for even Feagin RA. 2005. Artificial dunes created to protect property on
the colonizers to survive the stress. Galveston Island, Texas: the lessons learned. Ecol Restor 23:
The model produced graphical output which matched 89–94.
quite well with what has already been seen along the Gulf Feagin RA, Wu XB, Smeins FE, et al. 2005. Individual versus
Coast (Figures 1a–c) and the East Coast of the US community-level processes and pattern formation in a model
of sand dune plant succession. Ecol Model 183: 435–49.
(Roman and Nordstrom 1988); late-succession sand dune Gibeaut JC, Hepner TL, Waldinger R, et al. 2003. Geotubes for
species are being lost because embryonic dunes are not temporary erosion control and storm surge protection along
able to form in front of the barriers (Gibeaut et al. 2003), the Gulf of Mexico shoreline of Texas. Proceedings of the
leaving isolated “islands” of plant communities (Figure 13th Biennial Coastal Zone Conference; 13–17 July 2003;
2b) that are spatially removed from other communities, Baltimore, Maryland: NOAA. www.beg.utexas.edu/coastal/
hazard_atlas1.htm. Viewed 28 June 2005.
sometimes by considerable distances. Greipsson S. 2002. Coastal dunes. In: Perrow MR and Davy AJ
In the short term, money and time must be spent to (Eds). Handbook of ecological restoration. Volume 2:
restore these communities. In many cases, late-succession Restoration in practice. Cambridge, UK: Cambridge
species such as beach panic (Panicum amarum) have very University Press.
low rates of seed viability, and must therefore be planted Holgate SJ and Woodworth PL. 2004. Evidence for enhanced
coastal sea level rise during the 1990s. Geophys Res Lett 31:
during restoration efforts in order for them to appear L07305.
(Feagin 2005). The loss of such species is already occur- IPCC (Intergovernmental Panel on Climate Change). 2001.
ring on Galveston Island, where sea oats (Uniola panicu- Climate change 2001: impacts, adaptation, and vulnerability.
lata) has disappeared due to a combination of human- In: McCarthy JJ, Canziani OF, Leary NA, et al. (Eds).
induced disturbance and climate change (Greipsson Contribution of Working Group II to the Third Assesment
Report of the Intergovernmental Panel on Climate Change.
2002). In the long term, it may be critical to maintain New York, NY: Cambridge University Press.
propagule dispersal sites with planning done at a land- Leatherman SP. 1984. Coastal geomorphic responses to sea-level
scape scale in order to maintain populations along the rise: Galveston Bay, Texas. In: Barth MC and Titus JG (Eds).
East Coast of the US and in areas of large-scale develop- Greenhouse effect and sea-level rise. New York, NY: Van
ment, even for early succession species such as the endan- Nostrand Reinhold.
Leatherman SP, Zhang K, and Douglas BC. 2000. Sea level rise
gered sea-beach amaranth (Amaranthus pumilus). shown to drive coastal erosion. Eos 81: 55–57.
The maintenance of the late-succession species is criti- Martinez ML and Psuty NP. 2004. Coastal dunes: ecology and
cal, as they are usually the most important species in the conservation. New York, NY: Springer.
building of dunes, binding of sediments, and reduction of Miller L and Douglas BC. 2004. Mass and volume contributions
erosion. Moreover, these perennial species provide cover to twentieth-century global sea level rise. Nature 428:
406–09.
on the dunes throughout the year; if lost, erosion rates are
© The Ecological Society of America www.frontiersinecology.org
Loss of dune plant habitats RA Feagin et al.
Morton RA, Miller TL, and Moore LJ. 2004. National assessment Report 00–179. http://pubs.usgs.gov/dds/dds68/. Viewed 28
364 of shoreline change: Part 1, historical shoreline changes and June 2005.
associated coastal land loss along the US Gulf of Mexico. Titus JG, Park RA, Leatherman SP, et al. 1991. Greenhouse effect
USGS Open File Report 2004–1043. and sea level rise: the cost of holding back the sea. Coast
Morton RA, Pilkey Jr OH, Pilkey Sr OH, and Neal WJ. 1983. Manage 19: 171–204.
Living with the Texas shore. Durham, NC: Duke University Twilley RR, Barron EJ, Gholz HL, et al. 2001. Confronting climate
Press. change in the Gulf Coast region: prospects for sustaining our eco-
Nordstrom KF. 2004. Beaches and dunes of developed coasts. logical heritage. Cambridge, MA, and Washington, DC: Union
Cambridge, UK: Cambridge University Press. of Concerned Scientists and Ecological Society of America.
Pilkey OH and Frasier ME. 2003. A celebration of the world’s USGS. 2002. Houston–Galveston Bay area, Texas, from space – a
barrier islands. New York, NY: Columbia University Press. new tool for mapping land subsidence. Washington, DC:
Roman CT and Nordstrom KF. 1988. The effect of erosion rate on USGS. Fact Sheet 110–02.
vegetation patterns of an East Coast Barrier Island. Estuar Wolfram S. 1996. The Mathematica book. Cambridge, MA:
Coast Shelf S 26: 233–42. Wolfram Media/Cambridge University Press.
Stive MJF. 2004. How important is global warming for coastal Zhang K, Douglas BC, and Leatherman SP. 2002. Do storms cause
erosion? Climatic Change 64: 27–39. long-term beach erosion along the US East Barrier Coast? J
Thieler ER and Hammar-Klose ES. 2000. National assessment of Geol 110: 493–502.
coastal vulnerability to future sea-level rise: preliminary Zhang K, Douglas BC, and Leatherman SP. 2004. Global warming
results for the US Gulf of Mexico coast. USGS Open-File and coastal erosion. Climatic Change 64: 41–58.
www.frontiersinecology.org © The Ecological Society of America