Shorebird patches as fingerprints of fractal coastline fluctuations due to climate change

MISSING IMAGE

Material Information

Title:
Shorebird patches as fingerprints of fractal coastline fluctuations due to climate change
Physical Description:
Mixed Material
Language:
English
Creator:
Convertino, Matteo
Bockelie, Adam
Kiker, Gregory A.
Munoz-Carpena, Rafael
Linkov, Igor
Publisher:
BioMed Central (Ecological Processes)
Publication Date:

Notes

Abstract:
Introduction: The Florida coast is one of the most species-rich ecosystems in the world. This paper focuses on the sensitivity of the habitat of threatened and endangered shorebirds to sea level rise induced by climate change, and on the relationship of the habitat with the coastline evolution. We consider the resident Snowy Plover (Charadrius alexandrinus nivosus), and the migrant Piping Plover (Charadriusmelodus) and Red Knot (Calidris canutus) along the Gulf Coast of Mexico in Florida. Methods: We analyze and model the coupled dynamics of habitat patches of these imperiled shorebirds and of the shoreline geomorphology dictated by land cover change with consideration of the coastal wetlands. The land cover is modeled from 2006 to 2100 as a function of the A1B sea level rise scenario rescaled to 2 m. Using a maximum-entropy habitat suitability model and a set of macroecological criteria we delineate breeding and wintering patches for each year simulated. Results: Evidence of coupled ecogeomorphological dynamics was found by considering the fractal dimension of shorebird occurrence patterns and of the coastline. A scaling relationship between the fractal dimensions of the species patches and of the coastline was detected. The predicted power law of the patch size emerged from scale-free habitat patterns and was validated against 9 years of observations. We predict an overall 16% loss of the coastal landforms from inundation. Despite the changes in the coastline that cause habitat loss, fragmentation, and variations of patch connectivity, shorebirds self-organize by preserving a power-law distribution of the patch size in time. Yet, the probability of finding large patches is predicted to be smaller in 2100 than in 2006. The Piping Plover showed the highest fluctuation in the patch fractal dimension; thus, it is the species at greatest risk of decline. Conclusions: We propose a parsimonious modeling framework to capture macroscale ecogeomorphological patterns of coastal ecosystems. Our results suggest the potential use of the fractal dimension of a coastline as a fingerprint of climatic change effects on shoreline-dependent species. Thus, the fractal dimension is a potential metric to aid decision-makers in conservation interventions of species subjected to sea level rise or other anthropic stressors that affect their coastline habitat. Keywords: Land cover change, Coastal wetlands, Coastline complexity, Fractal dimension, Habitat suitability, Patches, Sea level rise
General Note:
Publication of this article was funded in part by the University of Florida Open-Access publishing Fund. In addition, requestors receiving funding through the UFOAP project are expected to submit a post-review, final draft of the article to UF's institutional repository, IR @ UF, (www.uflib.ufl.edu/UFir) at the time of funding. The institutional Repository at the University of Florida community, with research, news, outreach, and educational materials.
General Note:
Convertino et al. Ecological Processes 2012, 1:9 http://www.ecologicalprocesses.com/content/1/1/9; Pages 1-17
General Note:
doi:10.1186/2192-1709-1-9 Cite this article as: Convertino et al.: Shorebird patches as fingerprints of fractal coastline fluctuations due to climate change. Ecological Processes 2012 1:9.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
All rights reserved by the source institution.
System ID:
AA00013662:00001


This item is only available as the following downloads:


Full Text


Convertino et al. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


0 Ecological Processes
a SpringerOpen Journal


Shorebird patches as fingerprints of fractal

coastline fluctuations due to climate change

Matteo Convertino1,2,3*, Adam Bockelie2,4,5, Gregory A Kiker1,3, Rafael Munoz-Carpena1,3 and Igor Linkov6,7


Abstract
Introduction: The Florida coast is one of the most species-rich ecosystems in the world. This paper focuses on the
sensitivity of the habitat of threatened and endangered shorebirds to sea level rise induced by climate change, and on
the relationship of the habitat with the coastline evolution. We consider the resident Snowy Plover (Charadrius
alexandrinus nivosus), and the migrant Piping Plover (Charadrius melodus) and Red Knot (Calidris canutus) along the
Gulf Coast of Mexico in Florida.
Methods: We analyze and model the coupled dynamics of habitat patches of these imperiled shorebirds and of the
shoreline geomorphology dictated by land cover change with consideration of the coastal wetlands. The land cover is
modeled from 2006 to 2100 as a function of the Al B sea level rise scenario rescaled to 2 m. Using a maximum-entropy
habitat suitability model and a set of macroecological criteria we delineate breeding and wintering patches for each
year simulated.
Results: Evidence of coupled ecogeomorphological dynamics was found by considering the fractal dimension of
shorebird occurrence patterns and of the coastline. A scaling relationship between the fractal dimensions of the
species patches and of the coastline was detected. The predicted power law of the patch size emerged from
scale-free habitat patterns and was validated against 9 years of observations. We predict an overall 16% loss of the
coastal landforms from inundation. Despite the changes in the coastline that cause habitat loss, fragmentation, and
variations of patch connectivity, shorebirds self-organize by preserving a power-law distribution of the patch size in
time. Yet, the probability of finding large patches is predicted to be smaller in 2100 than in 2006. The Piping Plover
showed the highest fluctuation in the patch fractal dimension; thus, it is the species at greatest risk of decline.
Conclusions: We propose a parsimonious modeling framework to capture macroscale ecogeomorphological
patterns of coastal ecosystems. Our results suggest the potential use of the fractal dimension of a coastline as a
fingerprint of climatic change effects on shoreline-dependent species. Thus, the fractal dimension is a potential metric
to aid decision-makers in conservation interventions of species subjected to sea level rise or other anthropic stressors
that affect their coastline habitat.
Keywords: Land cover change, Coastal wetlands, Coastline complexity, Fractal dimension, Habitat suitability,
Patches, Sea level rise


*Correspondence: mconvertino@ufl.edu
1 Department of Agricultural and Biological Engineering- IFAS, University of
Florida, Gainesville, FL, USA
2Contractor at the Risk and Decision Science Team, Environmental Laboratory,
Engineer Research and Development Center, US Army Corps of Engineers,
Concord, MA, USA
Full list of author information is available at the end of the article

Springer 2012 Convertino et al, Icensee Springe
S^ rm er Attribution License (http://creativecom
S lin any medium, provided the original wo






Convertino et al. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


Introduction
Florida coastline-dependent species are characterized by
one of the highest extirpation risks in the world because
of sea level rise and increase in tropical cyclone activ-
ity (Convertino et al. 2010, 2011c) due to climate change.
The Snowy Plover (Charadrius alexandrinus nivosus;
SNPL hereafter) is a residential shorebird of Florida
listed as threatened at the state level. The Piping Plover
(Charadrius melodus; PIPL hereafter) is federally des-
ignated as threatened, and it migrates mostly from the
North Atlantic coasts of the USA and Canada to Florida
where it winters for 3 months on average (Elliott Smith
and Haig 2004). The Red Knot (Calidris canutus; REKN
hereafter) is designated as threatened in New Jersey and
is federally listed as a potential "at risk" species. REKN
uses the Florida Gulf beaches as stop-over areas for about
3 weeks during its migration between South America
and North America's Big Lakes region and Atlantic coast
(Harrington 2001). This is considered as the wintering
period of the REKN in Florida. An understanding of
the spatial distribution of the suitable habitat patches
for these shorebirds, their controlling factors, and how
these factors are affected by sea level rise is funda-
mentally important for adopting efficient conservation
strategies. An understanding of linkages between the cou-
pled evolution of landforms and ecological patterns is
a crucial topic due to the evidence that these patterns
are tightly linked. Biocomplexity approaches (Mandelbrot
1982; Rinaldo et al. 1995; Banavar et al. 2001; Pascual
et al. 2002; Schneider and Tella 2002; Buldyrev et al.
2003; del Barrio et al. 2006; Sol6 and Bascompte 2006;
Scanlon et al. 2007), despite being accused of adopting
simplified biological models (Paola and Leeder 2011), are
capable of reproducing macroscale patterns of complex
phenomena and of developing indicators, such as the
probability of the patch size (Mandelbrot 1982; Bonabeau
et al. 1999; Jovani and Tella 2007; K6fi et al. 2007; Jovani
et al. 2008; Convertino et al. 2012), that are useful for
assessing ecosystem health (Kefi et al. 2011). One of ecol-
ogy's main goals is to detect from observed patterns,
such as species occurrence patterns, the organizational
rules of species in stationary and evolving ecosystems.
Many theories have been proposed to explain the forma-
tion of clustered patterns of species in nature. Conspecific
attraction, environmental heterogeneities, and food avail-
ability have been claimed-alone or together-to be the
motivation for the formation of habitat patches in which
individuals of a species coexist in colonies. An optimal
search theory, the so-called L6vy-flight foraging hypoth-
esis (or predator-prey-food resource dynamics), predicts
that predators should adopt search strategies known as
L6vy flights where prey is sparse and distributed unpre-
dictably. However, Humphries et al. (2010) showed that
Brownian movement is sufficiently efficient for locating


abundant prey. This theory explains the clustered patterns
of resources in landscapes that may be different from the
pattern of species occurrence. Neither the L6vy-flight for-
aging hypothesis nor Brownian movement model address
the linkages of biota with landscape forms and their evo-
lution, which is, in our opinion, one of the main missing
points.
The colony size of seabirds (Schneider and Tella 2002),
colonial birds (Jovani and Tella 2007), and many other
other animals (Bonabeau et al. 1999) has been found
to follow a power-law distribution. Analogous scale-free
distributions have been detected for bacteria colonies
(Buldyrev et al. 2003), for species in complex ecosys-
tems (Sole and Bascompte 2006; Convertino et al. 2012),
and also for man-made systems such as cities (Batty and
Longley 1994). The ubiquity of the power-law structure
for the probability of the patch size in aggregation
phenomena of natural and human systems suggests
the existence of universal self-organization principles
(Pascual et al. 2002; Sole and Bascompte 2006). The scal-
ing exponent of the power-law distribution of the aggre-
gate size was proven to be the fractal dimension of the pat-
tern analyzed (Mandelbrot 1982; Convertino et al. 2012).
The word "aggregate" is a general word for indicating the
assemblage of individuals with similar or identical features
in a landscape. In the presence of a power law for the prob-
ability distribution of the aggregate size, the occurrence
patterns are scale-free, indicating that the patterns are
invariant at different scales of observations (Convertino
et al. 2012). The concept of fractal dimension was intro-
duced by Mandelbrot analyzing the coastline of Britain
at different scales (Mandelbrot 1967). The work dissem-
inated the use of fractal analysis first in g,-..!1.1i ..,i.1...-\
(Morais et al. 2011; Baldassarri et al. 2012) and later to
a variety of sciences from biology to engineering (Bak
1999). Nonetheless, all these theories, models, and empir-
ical findings have rarely considered any potential effect of
slow or abrupt change in the exogenous factors on the
heterogenous habitat in which species live. Only recently
it was proven quantitatively that ecosystems exhibit vari-
ations in the probability distribution of the patch size
due to anthropically and naturally driven changes in the
environmental variables (Kefi et al. 2011). For example,
desertification of water-controlled ecosystems produces a
decrease in the fractal dimension of vegetation patches, or
in extreme cases, a shift from the power law to exponen-
tial distribution of the patch size (K6fi et al. 2007; Scanlon
et al. 2007; Kefi et al. 2011). Climate change scenarios
tested in temperate/continental regions depicted an over-
all decrease in the fractal dimension of patches in time for
many different taxa (Barrio et al. 2006). For colonial birds
the variation in the fractal dimension of the patches was
clearly related to the fluctuations in the population abun-
dance due to interspecies competition (Jovani et al. 2008).


Page 2 of 17






Convertino etal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


In g>.. ,.i. i.!. 1..2\ the variation in the fractal dimension
was used as the signature of the persisting climate over
landscapes. For example, the association between land-
scape evolution and climate has been assessed for river
basin ecosystems in Rinaldo et al. (1995). However, none
of the previous studies linked the fractal dimension of two
ecosystems' patterns in time (e.g., of geomorphological
and ecological patterns) resulting from linked processes.
Here we verify for the first time, to the best of our knowl-
edge, that the fractality of the coastline is clearly linked to
the habitat patches of shoreline-dependent birds in their
breeding and wintering seasons.
We hypothesize that sea level rise may increase the
complexity of the coastline and that such complexity
determines fragmentation of the habitat of species. We
assume scale invariance of the patches, which is also
detectable by the analysis of the shorebird occurrences.
We consider a breeding shorebird (Snowy Plover) and
wintering shorebirds (Piping Plover and Red Knot) in
Florida to quantify the potential effect of sea level rise on
resident and migrant species. For the Snowy Plover the
nesting season is usually considered part of the breeding
season; thus, our model's input considers the SNPL breed-
ing and nesting occurrences simultaneously. Furthermore,
observations indicate that nesting, breeding, and winter-
ing areas for SNPL fall within the same range (Convertino
et al. 2011a). Wintering occurrences of SNPL are thus
considered together with breeding occurrences.
An integrated ecogeomorphological modeling approach
is adopted to predict the viability from 2006 to 2100
of threatened, endangered, and at risk (TER) shorebirds
(SNPL, PIPL, and REKN) along the Gulf Coast of Florida
as a function of the increasing sea level rise due to climate
change. We rescale to 2 m the Intergovernmental Panel
on Climate Change (IPCC A1B) scenario described in
Chu-Agor et al. (2011) and model the ecosystem at a 120
m spatial resolution. We predict land cover change with
the Sea Level Affecting Marshes Model [SLAMM (Clough
2010)] which is a geomorphological model at low-medium
level of complexity. SLAMM considers coastal wetland
types such as swamp, cypress swamp, mangrove, and salt
marsh (Additional file 1: Figure Sl). The habitat model
predicts the habitat suitability for breeding and win-
tering through a maximum entropy principle approach
(MAXENT) (Phillips and Miroslav 2008) as a function
of the recorded species occurrences in the breeding and
wintering season, the predicted land cover, and a geol-
ogy layer. MAXENT is an ecological model at low level
of complexity. The land cover and habitat simulations
are produced in Aiello-Lammens et al. (2011). Finally,
in this paper a patch-delineation model is introduced to
predict the yearly habitat patches for a set of biologi-
cal constraints imposed on the habitat suitability maps.
We assume the stationarity of the habitat patterns at


the year scale and absence of biological adaptation of
species to climate change. The fractal dimension of the
patches is derived by three independent methods: (i)
box-counting for the observed occurrences; (ii) probabil-
ity distribution of the patch size ["Kordak's law" (Korcak
1940; Mandelbrot 1982)]; and (iii) perimeter-area rela-
tionship for the predicted patches. We assume that these
three methods produce very close estimates of the frac-
tal dimension of the whole mosaic of patches as shown in
Convertino et al. (2012).
The power-law distribution of the patch size is verified
by almost a decade (2002-2010) of historical observations
of the species. Thus, the patch-delineation model is vali-
dated against these observations from 2002 to 2010. The
coupled ecogeomorphological organization is shown by
the correspondence in time of the fractal dimensions of
the habitat-specific coastline and of the predicted patches.
The fractal dimension of the habitat-specific coastline,
along with habitat loss and population abundance, is
demonstrated to greatly influence the number and size of
the patches, which are related to habitat loss and pop-
ulation abundance. Although the fragmentation of the
habitat (which is proportional to the fractal dimension
of the patches) is predicted to fluctuate considerably in
this century, the risk of extirpation of the species ana-
lyzed is not drastically increased because the connectivity
of the patches is predicted to increase. The Piping Plover
is the species with the largest fluctuation in the number
and size of patches. We believe the research presented
in this paper constitutes a contribution to the emerg-
ing field of biogeosciences, which explores the interface
between biology and the geosciences and attempts to
understand the interrelated functions of landscapes and
biological systems across multiple spatial and temporal
scales. We are aware of the existence of many other com-
plex ecogeomorphological processes that are not included
in our modeling effort. However, parsimonious models
such as the model presented here can capture large-scale
patterns while bypassing small-scale details (Ehrlich and
Levin 2005; Pascual et al. 2011). These models can be
tested against other more biologically realistic models to
fully explore the linkages among various environmen-
tal changes, geomorphological dynamics, and biodiversity
patterns. We anticipate that further research will explore
this issue of process complexity versus model complex-
ity, model relevance, and model uncertainty, which can be
synthesized as a "modeling trilemma" (Muller et al. 2010).
This paper is organized as follows. The "Methods"
section describes the shorebird data and the study site and
explains the models used in this study and the theoretical
characterization of patches. The "Results and discussion"
section reports the main results with a broad discussion
of figures and how these results are interpreted consider-
ing our assumptions. The "Conclusions" section reports


Page 3 of 17






Convertino et al. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


the most important conclusions, implications for manage-
ment, and further research efforts. Additional files 1 and 2
are provided to support our main result.

Methods
Site description and biogeographical variables
The white fine-sand beaches of the Florida coast of the
Gulf of Mexico constitute the habitat of the whole Florida
SNPL population. The SNPL population in Florida is dis-
tributed along about 80% of the Florida Panhandle and
along about 20% of the Florida Peninsula (Lamonte and
Douglass 2002; Himes et al. 2006; Burney 2009; Pruner
2010) (Figure la). The Florida Peninsula and the Atlantic
coasts are the main wintering grounds for the migra-
tory PIPL and REKN, which seem less constrained than
the SNPL by the mineralogical properties of the beach
substrate captured by the geology layer (Convertino et al.
2010,2011b). The land cover, which includes many wet-
land types from C-CAP (2009) is represented in Figure
S1 of the Additional file 1, and the geology (F-DEP
2001) characterizes the mineralogical substrate of each
land cover class (Additional file 1: Figure S6) (Convertino
et al. 2011b). In 2006 the PIPL Panhandle-Peninsula and
Atlantic populations were 38 and 33%, respectively, of
the total migrant PIPL population in Florida. The REKN
Panhandle-Peninsula and Atlantic populations were 55
and 20%, respectively, of the total migrant population
in Florida. The International Piping Plover Census in
2006 supported the field sampling of SNPL, PIPL, and
REKN (USGS-FWS 2009; FWC 2010; Alliance 2010).
The 2006 wintering occurrences in Florida are the data
used in this study for PIPL and REKN. For the SNPL,
data of breeding and nesting occurrences are also avail-
able from 2002 to 2010 and are provided year by year
by the Florida Wildlife Commission. These occurrences
are used to verify the assumption of scale-invariance
of SNPL occurrence patterns over time with the box-
counting. However, despite the availability of SNPL data
from 2002 to 2010, we construct the habitat suitability
model with the 2006 SNPL occurrences alone in order
to be consistent with the 2006 NOAA land cover (C-
CAP 2009) and the 2006 PIPL and REKN occurrences.
The geology and the elevation from USGS (USGS 2010;
Convertino et al. 2010, 2011b) are used in the habitat
suitability model and in the land cover model, respec-
tively (Aiello-Lammens et al. 2011; Convertino et al. 2010,
2011b).
We consider PIPL and REKN in the same geographic
domain where the full range of the SNPL occurs in order
to perform a simultaneous interspecies assessment of
the habitat use and extirpation risk of the three species
(Figure la). Thus, only the Panhandle-Peninsula region
was considered in this study. The SNPL is our main inter-
est because its year-round presence in the Florida coastal


ecosystem makes this species potentially more vulnera-
ble than PIPL and REKN. Dispersal among the Panhan-
dle and Peninsula SNPL populations has been observed
but not quantified. Population subdivision of the SNPL
has not been observed; thus, we can adopt the same
habitat and dispersal criteria for the whole population.
Population subdivision, for example, can be caused by
geographic barriers or disturbances [e.g., renourishment
(Convertino et al. 2011a)] that interfere with the disper-
sal. The reduction in dispersal is reported to reduce gene
flow and increase genetic drift of independent subpopu-
lations in the long-term. However, this is not the case for
the SNPL population in Florida despite the weak inter-
change of individuals between Panhandle and Peninsula
(Aiello-Lammens et al. 2011).
Habitat area and dispersal data for SNPL are mostly
from Aiello-Lammens et al. (2011) but also from Page
et al. (2009), Patons and Edwards (1996), Stenze et al.
(1994; 2007), Warriner et al. (1986). Aiello-Lammens et
al. 2011 synthesized the biological data and the metapop-
ulation modeling effort of this research for the SNPL.
Information is gathered also from field ecologists work-
ing on this project [i.e., Dr. R.A. Fischer (Engineering
Research and Development Center, US Army Corps of
Engineers) and Mrs. A. Pruner (Florida Park Service)]. For
PIPL, habitat and dispersal data are from Audubon (2006),
Seavey et al. 2010, and USFWS (2009), and for REKN, data
are from Fallon (2005) and Leyrer et al. (2006). For a more
detailed description of the site under study we refer the
reader to Convertino et al. (2011b).

Box-counting algorithm
The characterization of the occurrence patterns of breed-
ing and wintering occurrences and of the coastline is
performed using the "box-counting" method. For the
SNPL the occurrence pattern of nesting occurrences was
observed to be a self-similar pattern (Convertino et al.
2012); thus, the box-counting method is suitable to predict
how this pattern changes with the scale of analysis. The
box-counting analysis consists of calculating, for grids of
different box-side lengths, the number of boxes that con-
tain the object under study. Adjacent boxes constitute an
approximation of the real patches at each resolution. The
algorithm can be applied to both point and line patterns.
The box-counting is performed over eight orders of mag-
nitude in a logarithmic scale of the box-side length, from
lo(1) = 565 km (which corresponds to the box "Bl" in
Figure 1), which is approximatively the width of Florida,
to 1o(500oo) = 11.3 x 10-3 km (Figure 1). We indicate with
lo(i) the length of the box side at resolution "o=i" from
i = 1,..., 5000 where the increment from one resolution
to another is 11.3 x 10-2 km, which is slightly smaller
than the average home range of the SNPL (Table 1). The
order of magnitude is relative to the scales of analysis


Page 4 of 17







Convertino etal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


C-*-


/
0 29,500 59100 118,000 Meters
/


N


S.-.


N -
N. ****.


0 3,950 7,900 15,800 Meters
I I, I l l l ,


/.


0 100 200
1w 1-


~rnx
II.B4


Figure 1 Box-counting algorithm. (a) Representation of the the box-counting algorithm applied to the 2006 occurrences of the Snowy Plover
(SNPL), 11 i Plover (PIPL), and Red Knot (REKN), for eight orders of magnitude (in a logarithmic scale), which corresponds to 5000 resolutions of
the box-counting grid. In this example at the resolution of box B5 the number of boxes in which there is at least one occurrence is N(85) = 6. (b)
Box-counting example applied to the whole coastline, to the habitat-specific coastline (e.g., beach, salt-marsh), and to other land cover classes as in
Convertino et al. (2011 b). Many coastal wetland types are included in the land cover, such as swamp, cypress swamp, mangrove, and salt marsh. The
shaded grid cells in (a) and (b) have at least one species occurrence or a coastline segment at the represented resolution. Two coastline
configurations are presented: the first for high values of Dr and DK (c), the second for low values of Dr and DK (d). The patches presented in green
are connected because their neighboring distance is lower than the maximum dispersal length dc.


Page 5 of 17


H
$0.
/ \'
\


N
A


t E
s


I







Convertino etoal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


Table 1 Macroecological parameters of the
patch-delineation model, and biological data estimated
from the literature


Sp (kmi)
Sb/w (km2)
d( (km)


hr (km2)
hrd (km)
(ndi (km)
m (g)


0.03
0.01
8


0.016
0.12
0.72
38 40


Model parameters
0.04
0.02
12
Data
2.20
1.48 2.40
0.96
50 60


5p and Sblw are the minimum population and breeding/wintering area,
respectively, d is the estimated maximum dispersal length. {nd} is the
neighborhood distance in the breeding season for Snowy Plover (SNPL), and in
the wintering season for Piping Plover (PIPL) and Red Knot (REKN). hr and hrd are
the home range and the home range distance estimated considering the
breeding regions for SNPL, PIPL, and REKN. m is the average body mass. Data are
taken from Burney (2009), Himes et al. (2006), Lamonte and Douglass (2002),
Pruner (2010), Aiello-Lammens et al. (2011). Model parameters are assumed
considering data and calibrating the patch-delineation model on the observed
patch size in 2006 derived from the box-counting.


(extent) investigated by the box-counting, while the reso-
lution is related to the grids chosen for the box-counting.
The number N(1) of boxes of size I needed to cover the
pattern of occurrences (which is generally a fractal set)
follows a power law,

N(l) = No I-D, (1)

where D < d, and d is the dimension of the space (usually
d = 1, 2, 3). D is also known as the Minkowski-Bouligand
dimension, Kolmogorov capacity, or Kolmogorov dimen-
sion, or simply the box-counting dimension and is an esti-
mate of the Hausdorff dimension (Mandelbrot 1982). The
fractal dimension for 1 d objects is associated with the
Hurst exponent H such that D = 2-H (Mandelbrot 1982;
Bak 1999). The values of the Hurst exponent vary between
0 and 1, with higher values indicating a smoother trend,
less volatility, and less roughness of the analyzed pattern
(Mandelbrot 1982). We indicate the fractal dimension of
the breeding and wintering occurrences with Db and the
fractal dimension of the coastline with Df derived from
box-counting analysis. Both fractal dimensions are deter-
mined by the box-counting method. The fractal dimen-
sion of the coastline is calculated also for each land cover
class that is a species-specific habitat for the species con-
sidered (Figure ib). Many land cover classes are coastal
wetland types (Additional file 1: Figure Sl).

Land-cover model
The land cover is predicted year by year by using the Sea
Level Affecting Marshes Model (SLAMM) (Clough 2006;


Chu-Agor et al. 2011) starting from the year 2006 to 2100.
These simulations are performed in Aiello-Lammens et al.
(2011) and Convertino et al. (2010) to which we refer
the reader for more details. The domain of the model
is extended inland for about 10 km from the coastline
(Convertino et al. 2011a,2011b) (black region along the
coast in Figure 1, box Bl). We consider the predicted inun-
dation distance in 2100 (~ 9 km) for a range of [1, 2] m sea
level rise (SLR) adding 1 km to consider the uncertainty
in the estimation of the flooding distance. The initial
condition is the 2006 land cover from NOAA (Klemas
et al. 1993). The NOAA land cover classes are changed
into SLAMM land cover classes for modeling purposes.
SLAMM requires us to group the classes of land cover into
model classes. The conversion is reported in Convertino
et al. (2011b). The SLAMM model also requires the ele-
vation and slope as input variables. The modeled domain
is divided into seven regions (Additional file 1: Figure
Sl) with distinct historical tidal and SLR trends. Each
region is characterized by a unique set of values for the
26 input parameters (Additional file 1: Table Sl) related to
tide, accretion, sedimentation, and erosion processes. The
value of the parameters is derived from the available liter-
ature and previous efforts of this research (Chu-Agor et al.
2011). In this effort of modeling the land cover, we do not
consider any geomorphological feedback between land-
forms and climate change that is expected to occur with
global warming. All our assumptions are the same as those
in Chu-Agor et al. (2011) and Convertino et al. (2010).
Also we do not consider any possible barrier island shift-
ing because that is reported to occur over a time period
much longer than our predictions (Masetti et al. 2008).

Habitat suitability model
The employed habitat suitability model is MAXENT
(Phillips et al. 2006; Phillips and Miroslav 2008), which is
one of the most diffused models in species distribution
modeling. MAXENT is a model based on the principle of
maximum entropy that predicts continuous habitat suit-
ability maps of potential species occurrence under a set
of selected environmental variables. The environmental
variables that are necessary and sufficient for calculating
the habitat suitability are the land cover translated into
SLAMM classes (Chu-Agor et al. 2011; Convertino et al.
2011a,2011b) and the USGS geology layer (Convertino
et al. 2011a,2011b) at a resolution of 120 m. The resolution
120 m is the home-range distance of the SNPL (Table 1).
Such distance is sufficient to capture not only the spatial
variability of habitat preferences of SNPL, but also that of
PIPL and REKN, whose home-range distance are much
larger than that of SNPL. The habitat suitability at-a-point
(i.e., for each pixel of the modeled domain) can be con-
sidered as a proxy to find SNPL, PIPL, and REKN in the
breeding and wintering season. The prior probabilities of


Page 6 of 17






Convertino etal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


occurrence are calculated in MAXENT using the recorded
shorebird occurrences constrained to the environmental
variables. The occurrences are nest and breeding occur-
rences for SNPL, and the adult occurrences for PIPL and
REKN. Thus, for PIPL and REKN the habitat suitabil-
ity refers to the suitability for wintering as in Convertino
et al. (2011a). No absences are required in MAXENT.
Then the posterior probabilities of occurrence are based
on the prior probabilities given the change in the land
cover modeled year by year by the land cover model. A
regularization parameter that controls the fit of the pre-
dicted suitability to the real occurrence data is assumed
to be equal to one. Non-randomly placed pseudoabsences
are used to improve the predictions, and 25% of the occur-
rences are taken as a training sample (Convertino et al.
2011a,2011b). The predicted habitat suitability maps rep-
resent the average of over 30 replicates for each year to
reduce the uncertainty of the predictions. The habitat
suitability is calculated with 10,000 random background
points. Background points are a subset of points of the
domain over which the Bayesian inference between the
recorded species occurrences, pseudoabsences, and envi-
ronmental layers is determined.
We assign a biological interpretation to the predicted
habitat suitability score, P(hs), which is the probability at-
a-point of finding a breeding and/or a wintering ground.
Breeding and wintering grounds are suitable sites for the
SNPL as a function of the season considered, and winter-
ing grounds are suitable sites for the PIPL and REKN. We
define the suitability index (SI) as a metric from 0 to 100
that captures the quality of the breeding and/or wintering
habitat for the species. The higher the SI the larger the bio-
logical spectrum of functions performed by the species in
that habitat. Hence, P(hs) is also a surrogate of habitat use
during the breeding and wintering seasons of the species
considered. In fact, it is legitimate to assume that habi-
tat use increases with habitat quality. Every pixel of the
HS maps is classified into five SI categories: SI=100 [for
0.8 < P(hs) < 1] is considered the best habitat with the
highest survival and/or reproductive success; SI=80 [for
0.6 < P(hs) < 0.8] is typically associated with successful
breeding and/or wintering; SI=60 [for 0.2 < P(hs) < 0.6]
is associated with consistent use for breeding and winter-
ing; SI=30 [for 0.2 < P(hs)] is associated with occasional
use for non-breeding, feeding activities, and wintering; all
values less than SI=30 indicate habitat avoided both for
breeding and wintering; and SI=0 for completely unsuit-
able habitat. We refer the reader to Convertino et al.
(2011a) for additional details about MAXENT runs for the
SNPL, PIPL and REKN.

Patch-delineation model
Below we define a criterion for delineating breeding and
wintering patches for SNPL, PIPL and REKN, respectively.


A patch is defined when the following criteria simultane-
ously hold:


for


SI > 60 [i.e., for P(hs) > 0.2]
V P(X) within a neighborhood distance nd < dl
Minimum population patch size > Sp
Minimum breeding/wintering patch size > Sb/w


(2)

The species-dependent values for the three parame-
ters required for the patch identification are reported in
Table 1. The values of biological data in Table 1 are used
only to support the choice of model parameters. The
model parameters are calibrated to reproduce a patch-size
distribution as close as possible to the box-counting dis-
tribution of occurrences in 2006. The model with this set
of parameters was validated against the patch-size distri-
butions from 2002 to 2010 estimated by the box-counting.
We define breeding patch as an area large enough to at
least occasionally support a single breeding pair through
courtship and rearing of young to dispersal age (Majka
et al. 2007). A population patch is defined as an area large
enough to support breeding for 10 years or more, even if
the patch is isolated from interaction with other popula-
tions of the species (Majka et al. 2007). Since population-
wide data are lacking for these breeding and population
area requirements, we assumed that a population patch
is at least two times larger than a breeding patch. For
the SNPL these patches contain certain nesting patches.
The minimum population and breeding/wintering patch
areas are estimated from the literature available and
by expert knowledge of the field biologists involved in
the sampling campaigns performed for this study (see
Burney 2009; Himes 2006; Lamonte and Douglass 2002;
Pruner 2010). Sp and Sb/w are the minimum population
and breeding/wintering area, respectively, and are pro-
portional to the estimated home range. The minimum
breeding/wintering area is the minimum area that will
support breeding and wintering activity of the shorebirds.
The home range hr and the home-range distance hrd (the
square root of the hr) are values estimated considering the
breeding regions for SNPL, PIPL, and REKN. We assume
that Sp and Sb/w for PIPL, and REKN are much smaller
than hr because they refer to the wintering period of these
shorebirds in Florida. For REKN, Sp is also reduced due to
the habitat limitation and the close coexistence with SNPL
in the same habitat. Patches are considered connected
if their neighboring distance is equal to or smaller than
di, which is the maximum dispersal length. Figure lc,d
shows an example of patches that are connected because
their reciprocal distance is lower than d[. These plots
also represent our assumption that coastline complexity
affects patch distribution. The average neighborhood dis-
tance (nd) is the average dispersal of the species. (nd) is


Page 7 of 17







Convertino et ol Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


higher than hrd for the SNPL due to the higher local dis-
persal ability estimated from recent surveys (Himes et al.
2006; Pruner 2010). For PIPL and REKN, (nd) is smaller
than hrd because the reported hrd refers to their breed-
ing range in northern states in the USA and Canada. In
the winter season PIPL and REKN migrate to Florida,
and their dispersal distance is observed to be smaller.
Within the neighborhood distance a subpopulation can
be assumed to be panmictic. A panmictic population is
one in which all individuals are potential partners. It is
usually estimated from the foraging distance of an animal
species. In a more abstract way the neighborhood distance
is the glue of all the suitable patches. In a particle physics
analogy, it describes the Brownian motion of individuals
within a larger species group. Thus, by using di, which is
the maximum dispersal, as a criterion in the model, forag-
ing is certain to be considered within patches. Our model
considers an upper estimate of the patch size for all the
shorebirds considered. m is the average body mass, which
is used to discuss some results. We assume the same bio-
logical parameters for the SNPL Panhandle and Peninsula
as in Aiello-Lammens et al. (2011).

Probability distribution of the patch size
The probability of exceedance of the patch size is known
in literature as Kordak's law (Korcak 1940; Nikora et al.
1999), which is expressed by:

P(S > s) = cS- F (3)

where c is a constant, F is a homogeneity function that
depends on a characteristic size sc, and c = DK/2 is the
scaling exponent (Korcak 1940; Mandelbrot 1982). DK is
the fractal dimension of the patches. The probability of
exceedance exhibits a power-law behavior. The probability
distribution of the patch size for the predicted patches was
used to validate the patch-delineation model against the
box-counting estimates on the real occurrences from 2002
to 2010. The fit of the predicted distribution of patches is
performed using a Maximum Likelihood Estimation tech-
nique (MLE), which is described in the Additional file 1.

Perimeter-area relationship
The scaling relationship between the perimeter p and the
size S of the patches:

p = k SDc/2, (4)

determines the fractal dimension of the mosaic of patches,
which considers the fractality of the patch edge. Here
we indicate the fractal dimension Dc, which is derived
from the same predicted patches of the introduced patch
model (see the "Patch-delineation model" section) but
also considers their perimeters. Because Kordak's law
(Korcak 1940) considers only the size of the patches, the


perimeter-area scaling law has been considered as a more
precise tool for measuring the fractal dimension. In liter-
ature the ratio p/S is adopted to measure the quality of
the patches for population survivability, that is, the likeli-
hood of surviving in a suitable patch (Helzer and Jelinski
1999; Airoldi 2003; Imre and Bogaert 2004). In general the
higher the ratio p/S the less suitable the patch area for the
species, and the higher the ratio p/S, the higher the fractal
dimension Dc.

Results and discussion
The relationship between the number of cells occupied by
shorebird occurrences, N(1), and the length of the side of
the box, 1, at each scale of analysis is shown in Figure 2.


10 10 10 o' 110

Figure 2 Box-counting scaling-law in time. (a) Power law
N(1) = No /-DB derived from the box-counting algorithm applied to
the occurrences of PIPL (black dots) and REKN (green) in 2006 and to
the whole Florida Gulf coastline. In the inset the schematized Florida
coastline is evaluated at different box sizes. (b) Box-counting
algorithm applied to the 2002-2010 occurrence of the SNPL. The
fractal dimension derived from the analysis of the breeding and
nesting occurrences is Db = 1.63, 1.62, 1.75, 1.74, 1.63, 1.64, 1.66, 1.68,
1.70 for the years from 2002 to 2010, respectively. In the inset Db and
Df are reported for each year. Values of Df are reported (Additional file
1: Table S3).


Page 8 of 17


10 10' 10


10o 10







Convertino et al. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


The relationship is a power-law function, N(I) ~ lDb,
whose exponent Db is the fractal dimension of the shore-
bird occurrence pattern. Figure 2a reports the power-law
relationship for PIPL and REKN breeding occurrences in
2006, and Figure 2b for the SNPL breeding occurrences
from 2002 to 2010. The results confirm the supposed
scale-free distribution of the shorebird occurrences. The
fractal distribution of the predicted patches is captured by
Korcak's law (Figure 3). The box-counting overestimates
the fractal dimension with respect to the fractal dimen-
sion of Korcak's law as shown in Convertino et al. (2012).
The fractal dimension of the box-counting (Db) is 1.63,
1.85, and 1.53, and the fractal dimension of Korcak's law
(DK) is 1.47, 1.70, and 1.42 for SNPL, PIPL, and REKN
in 2006, respectively (Additional file 1: Tables S2 and S3).
The box-counting envisions a more pessimistic scenario
for the patch size of shorebirds. However, as in Convertino
et al. (2012) we believe that in the absence of any modeling
effort box-counting constitutes a valid technique to cal-
culate the fractal dimension of the mosaic of patches. For
the SNPL occurrences, box-counting allows us to detect
the fluctuation over time of the fractal dimension of the
recorded nest occurrences and of the coastline. The insets
in Figure 2b show the empirical evidence of the correla-
tion between Df and Db, and Additional file 1: Table S3
reports the values of the fractal dimensions. The anal-
ysis raises the question of whether the variation in Db
is caused by natural fluctuations of the species range or
by changes in external forcing such as natural or anthro-
pogenic stressors. We observe that in 2004 and 2005 the
fractal dimension showed a jump possibly due to the
exceptional hurricane season in those years, which altered
the positive feedback between tropical cyclones and SNPL
nest abundance (Convertino 2011c). This supposition is
confirmed by the results of Convertino et al. (2011c).
The potential effect of sea level rise, one of the main
controlling factors of land cover of coastal habitats, is
studied here. The simulated variation in land cover classes
over time is performed in SLAMM (Clough 2010) for
the Gulf Coast of Florida (Additional file 1: Figures S1
and S2). We predict by 2100 a decrease in the salt-marsh
and estuarine beach classes, which are crucial habitats for
PIPL, SNPL, and REKN. We also predict a net decrease in
swamp and inland fresh marsh habitats. Following flood-
ing predicted to occur after 2060, undeveloped drylands
will change mostly into tidal flats, which may shift into
estuarine open water (Additional file 1: Figure S2). We
estimate a 6% increase in estuarine open water and a
10% increase in ocean open water from 2006 to 2100.
We expect global land-loss independent of the land cover
class of about 16% with respect to the 2 m sea level rise.
A video in Additional file 2 and Figure S2 in Additional
file 1 show the evolution of land cover and of the coast-
line g.. il...l....\ over time. Additional file 1: Figures


o2006 o2020 o2040 02060 2080 o2100


10 10 10 10


10, 10 10 106 10


100 102 104 106 10,
S (nm)
Figure 3 Korcak's law of the predicted suitable patches. The
fractal dimension of the patches is derived from the scaling exponent,
c = DK/2, of the probability of exceedance of the patch size
(Equation 3) for SNPL, PIPL, and REKN. The probability of exceedence
of the patch size is represented for the years 2006, 2020, 2040, 2060,
2080, and 2100. The probability ofexceedance is compared against
the box-counting scaling laws for the years 2002-2010. Additional file
1: Table S2 reports the values of DK. The insets represent the
probability density functions of the patch size that show a
heavy-tailed behavior.


S3, S4, and S5 report the suitability index derived from the
predicted habitat suitability maps using MAXENT corre-
sponding to the yearly land cover maps. The patches are
then calculated using the patch-delineation model intro-
duced in the "Patch-delineation model" section and the
habitat suitability maps.


Page 9 of 17







Convertino etoal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


The power-law structure of the patch size holds for
every year simulated (Figure 3), which proves the scale-
invariance of the suitable habitat over time. By using the
maximum likelihood estimation (MLE) criteria, we found
that the Pareto-L6vy probability function has the best fit
for the predicted distribution of the patch size (Additional
file 1). Kordak's law exhibits some finite-size effects before
the upper truncation and a potential lower-cutoff in the
power-law behavior. However, these variations from the
power law are quite common in natural systems due to
the finiteness of the variable sampled. Thus, we can claim
an overall scale-invariance of the patch size. Additional
file 1: Table S2 reports the fractal dimension derived
from Kordak's law for 2006, 2020, 2040, 2060, 2080, and
2100. The scale-invariance of the habitat patterns of the
SNPL was shown in Convertino et al. (2011b) for the pre-
diction of the habitat suitability in 2006. Here we show
that, given the scale-invariance of the patch size, fluctua-
tions in the scaling exponent c = DK/2 of Kordak's law
occur. We believe that these fluctuations are related to
variations in the land cover, which changes the coastline
fractality. The higher the fractal dimension, the higher the
fragmentation of the shorebird habitat. The fragmenta-
tion of the habitat creates smaller patches for wintering
and breeding for PIPL and REKN, and for SNPL, respec-
tively. Brownian-L6vy movements of shorebirds might be
the cause for the scale-invariance of the occurrence pat-
terns that can be detected by the box-counting. This
has been proven for other marine animals (Humphries
et al. 2010) and colonial birds (Jovani et al. 2008). How-
ever, in this study we do not reproduce any movement
of species as we believe that the size and number of
patches is affected by the geomorphological evolution
of the coastline, which in turn affects the movement
of shorebirds.
The worst scenario for the vulnerability of shorebirds
is predicted considering the fractal dimension Db from
the box-counting. Moreover, the box-counting suffers
from the risk of potentially unsampled occurrences. The
Kordak's law fractal dimension (Figure 3) is based on the
size of the predicted suitable patches (\ .i,.,li., habitat
range), while the box-counting (Figure 2) is an approxi-
mation that only captures the recorded occurrences (real-
ized range). The fact that DK j Db for the 2006-2010
period in which SNPL nest occurrences are available con-
firms the good estimation of MAXENT of the realized
range as previously found in Convertino et al. (2011b).
A more accurate estimation of the fractal dimension
that is intermediate between DK and Db is given by the
patch perimeter-size scaling relationship (Figure 4). The
perimeter-size relationship captures the edge effects of
patches on species. In general shorebird species prefer to
live in patches whose shapes are as regular as possible ver-
sus highly irregularly shaped patches such as the patches


100




10



1




0.1
0.01
100 -




10 .


* 2100 SP
* 2006 SP


* 2100 PP
* 2006 PP


0.90 7.


. 0.875


0.1 1-
0.01
100 -




10 -1


I 2006 RK











c)


0.1 I
0.01 0.1 1 10
S (krn)
Figure 4 Perimeter-size relationship for SNPL, PIPL, and REKN.
Perimeter-size relationship (p = k SOD1/2) for the predicted suitable
patches of the SNPL (a), PIPL (b), and REKN (c), in 2006 and 2100. The
exponent Dc for the SNPL is listed in Additional file 1: Table S3.



determined by a very complex coastline. The survivabil-
ity of the species is higher for those inhabiting patches
with large perimeters and simple shapes than for those
inhabiting patches of equivalent area but complex shape.
The larger the edge effect determined by the complex-
ity of the patch parameter, the lower the probability of
survival for the individuals within the species within the


Page 10of 17


0.81 .


0 .
*, % .* .. 07 .... ..







Convertino etal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


patch. However, there are some cases of "edge species" for
whom irregular shapes are preferred. In our case it was
observed that DK < Dc < Db. Hence, the estimation of the
fractal dimension by using Kordak's law forecasts the best
scenario predicting the least amount of fragmentation due
to sea level rise. Dc predicts greater fragmentation than
DK because the fractality of the patch's perimeter is con-
sidered, but overall Dc seems the best estimate of the frac-
tal dimension between Kordak's law and the box-counting
estimates.
Figure 5a,b respectively, show the time series of the frac-
tal dimension of the species-dependent habitat coastline
Df (mostly beach for SNPL, PIPL, and REKN, but also
salt marsh for the PIPL), and of the fractal dimension
of the patches DK (from Equation 3) computed with the
patch-delineation model. Additional file 1: Figures S3, S4,
and S5 show the patches for SNPL, PIPL, and REKN in
the years 2006, 2020, 2040, 2060, 2080, and 2100. The
majority of patches are along the barrier islands and par-
ticularly in the Panhandle region. After 2060, when sea
levels start to rapidly rise, a consistent portion of the
patches will be found along the shore as barrier islands
gradually disappear. Figure 5a also shows the variation
in the fractal dimension of the whole coastline indepen-
dently of the land cover class. The probability of finding
a patch of size S is lower in 2100 than in 2006. DK values
are similar for SNPL and REKN and are higher for PIPL
(Figure 5b). Thus, on average the relationship DcRE..v
DKSNPFL < DjKPIPL holds for the modeled period. Big varia-
tions in DKPjPL. are observed particularly in correspon-
dence with big variations in the salt-marsh habitat
(Figure 5a,b), which confirms the likelihood of finding a
breeding ground of PIPL in the salt-marsh habitat (class
8 contained in Additional file 1: Figure S6) as reported in
literature (Convertino et al. 2011a) and as found by our
results (Additional file 1: Figure S6). In 2100 the fractal
dimension of REKN is very similar to the fractal dimen-
sion of SNPL, while the fractal dimension of PIPL is the
highest. The PIPL shows the lowest probability of large
patches with respect to the other shorebirds considered
because DK is the highest. The area under the power-
law distribution of patches for 2100 in Figure 3 has a
5, 3 and 8% negative variation with respect to the area
for 2006 for SNPL, PIPL and REKN. The area under the
curve is the overall probability of finding patches of any
size in a given year. Just comparing 2006 with 2100 is
not enough to derive any conclusion about the species
with the highest potential risk of decline. The location and
size of the patches are determined by the habitat suitabil-
ity at-a-point and by a combination of dispersal and area
criteria (see the "Patch-delineation model" section). The
Piping Plover, despite having a larger spectrum of habi-
tat preferences than SNPL and REKN [transitional marsh
and salt marsh areas are favorable classes as shown in


1.20 -0-
... 0oa
1.18


1.14 .
1.12
1.10
1.08 /6
1.06 ..
1.04
1.02
1.00
2000 2010
1.86 ---

1.76

1.66

1.56

1.46

1.36 .

1.26

1.16 ..

1.06


2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100
year


* SP RK E PP


1.8



1.6 -

r) 1.5-


1.06 1.08 1.10 1.12 1.14 1.16
Df


1.18 1.20


Figure 5 Fractal dimension time series of the shorebirds patches
and of the coastline. (a) Time series of the fractal dimension Df of
the entire coastline (blue line), of the salt-marsh (red), and of the
beach (green) habitat coastlines, determined by the box-counting
algorithm. (b) Fractal dimension DK over time for the patches for
SNPL (blue dots), PIPL (red), and REKN (green) derived from Korcak's
law. The dashed gray lines (a, b) represent the 95% confidence
interval of the estimated Df and DK. (c) Scaling relationship among
the fractal dimension of the patches for the threatened, endangered,
and potentially at-risk shorebird species (TER-s) and the fractal
dimension of the favorable habitat coastlines (salt marsh for PIPL, and
beach for SNPL and REKN). The average species-independent scaling
exponent is y= 1.67. The gray cloud (c) represents the 95%
confidence interval for the linear regression between DK and Dr.


Page 11 of 17


2020 2030 2040 2050 2060 2070 2080 2090 2100







Convertino etoal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


Additional file 1: Figure S6 and as reported in Convertino
et al. (2011a)] seems to be at risk due to the high frag-
mentation of its habitat. This is evidenced by the larger
fluctuation of DKPIpL than DKSpL and DKREKN.
We believe that it is important to observe the fluc-
tuations of DK over time for each species. DK values
of SNPL and REKN are on average steady and increas-
ing over time, respectively; thus, the probability of find-
ing large patches for these shorebirds decreases over
time with respect to 2006. DK of PIPL has the largest
fluctuations, but most of these fluctuations imply an
increase in the probability of finding large patches with
respect to 2006. Nonetheless we believe the frequent
and large variation in patches is not a good scenario
for species.
In Figure 5c we propose a scaling relationship between
the fractal dimension of patches and the fractal dimen-
sion of the habitat-specific coastline, DK ~ DfY. The
relationship holds over at least two orders of magnitude,
from the smallest patches (~ 0.01 km2) and short coast-
line segments to the largest patches and the whole Florida
Gulf coastline. The same scaling exponent is observed
for SNPL, PIPL, and REKN, underlining a possible com-
mon ecogeomorphological organization of the landscape
under sea level rise pressure. In Figure 5c Df is char-
acteristic of the portion of coastline in which there is
a suitable habitat for SNPL, PIPL, and REKN, which is
evidenced in Additional file 1: Figure S6. The coupled evo-
lution of the land cover and habitat patterns may hold
clues about the linkage of geomorphological and ecologi-
cal processes. The scaling relationship between the fractal
dimensions of patches and coastline can be a potential
tool to measure the vulnerability of the species in the
future. The higher the exponent y, the higher the poten-
tial risk of decline of the species. For small changes in
the configuration of the coastline, a large fragmentation
of the suitable habitat would potentially be observed. For
species with comparable values of y, which is the case
for SNPL, PIPL, and REKN, the range of values of DK
and Df is important for detecting which species may be
subjected to the most significant change in the suitable
habitat patches. The lower DK, the higher the likelihood
of having large patches. To the best of our knowledge this
is the first scaling relationship to be identified between
fractal dimensions of landscape and ecological patterns.
In this respect this relationship brings insights into the
field of "landscape allometry," which is the study of the
possible scaling of landscape and ecological patterns and
processes. The relationship is between fractal dimen-
sions, which are indicators that focus on how measured
quantities vary as a power of measurement scale, but at
the same time the relationship has an allometric focus,
between the coastline complexity and the magnitude of
habitat fragmentation.


However, fragmentation per se does not directly imply
loss of connectivity among patches. Figure 6 shows how
the average size of the patches (s) for SNPL, PIPL, and
REKN decreases with the increase in the fractal dimen-
sion of the patches. Here we consider DK of Kordak's law
for the fractal dimension. At the same time we observe


S6

4

2

0
1

1200

1000

800

2 600

400

200

0


SP PP RK

,' a)
\

S. mm
.. U
* N **E *


.2 1.3 1.4 1.5 1.6 1.7 1.
DK


U


* *~


1.2 1.3 1.4 1.5
DK


.


/
* -
Vi -


1.6 1.7 1.8


251
c)
20 -


15
/
'/ m "m I
10 / -
'
5" /
/
/

1.2 1.3 1.4 1.5 1.6 1.7 1.8
DK

Figure 6 Relationships among patch number, size, and
connectivity, and fractal dimension of the habitat-specific
coastline. (s) vs DK (a), Np vs DK (b), Np vs (s) (c), and (c) vs DK (d) for
the threatened, endangered, and at-risk shorebirds (TERs) considered.
The dots are the bin averages over 30 simulations for each year for
the period 2006-2100. The dashed lines represent the 95%
confidence intervals for the dependent variables considered.


Page 12 of 17






Convertino etal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


an increase in the number of patches Np. Thus, the
variation in the coastline produces fragmentation, rather
than shrinking, of the suitable habitat. The former does
not imply the latter as erroneously assumed by many
theoretical models in the ecological literature. The aver-
age size of the PIPL patches is lower than that for SNPL
and REKN, and the habitat for the PIPL is the most frag-
mented (Np is the highest on average). This is related
to the high value of DK for the PIPL with respect to
SNPL and REKN. Thus, although the variations in DKpp,
would predict bigger patches, the fragmentation of the
PIPL habitat is the greatest. In 2100 the number of suit-
able patches for SNPL, PIPL, and REKN is predicted to be
higher than in 2006, but the average size of the patches
is predicted to be smaller (Additional file 1: Table S2). As
sea level rise (SLR) increases the complexity of the coast-
line, habitat patches moderately shrink and split. On the
contrary when the coastline complexity decreases, habitat
patches enlarge and coalesce (Figure ic) as in our assump-
tion depicted in Figure lb. The PIPL seems to be the
shorebird most affected by the changes in its breeding
habitat due to sea level rise.
The average size and the number of the patches
are inversely proportional given the relationship in
Figure 6a,b and as shown in Additional file 1: Figure S7.
The average patch size (s) for the shorebirds is not pro-
portional to the average body mass m as possibly expected
(Table 1), although the latter scales with the average dis-
persal length. The (s) is for the PIPL, while it is larger for
SNPL and REKN. This emphasizes the controlling role of
habitat g,.. *in. **1....\ in shaping the patch distribution.
The PIPL also depends on the salt-marsh habitat, which t
is one of the classes more seriously compromised by SLR.
We consider dl, the estimated maximum dispersal length,
in order to determine the average number of connected
patches (c). dl considers rare "Levy flights" of individuals
of the species in the ecosystem. Levy flights are a spe-
cial class of random walk with movement displacements
drawn from a probability distribution with a power-law
tail (the so-called Pareto-Levy distribution), and they
give rise to stochastic processes closely linked to fractal
geometry and anomalous diffusion phenomena. Because
it has the largest maximum dispersal distance, the REKN
has the highest number of connected patches. However,
for the three shorebird species (c) increases with the
fractal dimension of the patches, indicating a measure of
the habitat fragmentation. Because we find that climate
change is responsible for the splitting of the patches,
rather than their shrinking, and because the dispersal
capability of species is not expected to change consistently
in the modeled period, the result seems justifiable. The
increase in the number of connected patches is explain-
able because Np increases without a drastic reduction
in the habitat. The average connectivity of the predicted


breeding and wintering patches is an increasing function
of the fractal dimension of the patches. The increas-
ing roughness of the Florida coastline due to climate
change produces a larger number of patches with smaller
dimensions. The increased connectivity would potentially
enhance the survivability of the shorebirds despite the
decrease in the average size of suitable patches. Thus, the
predicted patch patterns for the Florida shorebirds are
not the worst case scenario in which both the connectiv-
ity and the dimension of the patches are reduced. Further
explanation of the land cover, habitat, and patch dynamics
is provided in Additional file 1.

Conclusions
Sea level rise due to climate change, beyond being a
human-population threat, is shown to strongly affect
biodiversity such as residential and migrant shorebird
populations in Florida. The integrated patch-prediction
modeling framework proposed in this paper constitutes a
parsimonious but useful risk assessment tool for species
decline with respect to more accurate metapopulation
models. In our opinion, the understanding of ecogeomor-
phological processes at any scale of analysis together with
the detection of useful indicators of such dynamics is
one of the primary goals to protect biodiversity against
the anticipated changes in the landscape due to climate
change. On the one hand, it is impossible to consider, or
to estimate with low uncertainty, all the factors affecting
the processes that govern the distribution of species (e.g.,
conspecific attractions, interspecific competition, density
dependence, sex structure, life history, phenotypic plas-
ticity, and phenological changes in dispersal ability and
in breeding/wintering area requirements), the geomor-
phological processes, and the links and feedbacks among
these processes. On the other hand, we believe that a
top-down approach of biocomplexity is useful to detect
the fundamental drivers of the observed patterns of inter-
est (Schwimmer 2008; National Research Council 2009;
Reinhardt et al. 2010). We are aware that many geomor-
phological and biological processes are not incorporated
in the presented model; however, the uncertainty in the
quantification of these processes and the interaction of
these uncertainties may produce erroneous results in the
predictions. The integrated model is capable of provid-
ing valuable macroscale predictions with relatively few
data and variables. Thus, the model is useful for evalu-
ating conservation actions for increasing the survivability
of shorebirds in Florida. We are also confident that the
proposed model, properly tuned, can be applied to many
different species in coastal ecosystems worldwide that
are threatened by sea level rise. We anticipate further
development of this model at higher levels of complex-
ity and also for inland sites. The following conclusions are
worth mentioning.


Page 13 of 17






Convertino et al. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


* A scale-free distribution of nesting, breeding, and
wintering occurrences was detected for the Snowy
Plover in Florida. The scale-free distribution was also
found for the wintering occurrences of Piping Plover
and Red Knot. The distribution was derived through
the box-counting technique applied to the breeding
and wintering occurrences, which gives a proxy of the
fractal dimension of shorebird patches. Empirical
evidence shows that the fractal dimension of the
occurrences is strongly positively correlated with the
coastline fractal dimension, which underlines an
ecogeomorphological organization, i.e., a coupling of
ecological and geomorphological patterns. The
power law held for any season of the shorebird
annual cycle, demonstrating the high importance of
the physical habitat on species processes.
* We predicted breeding and wintering patches of
shorebirds, simulating land cover (which comprises
many coastal wetland types) and habitat suitability at
the year scale from 2006 to 2100 as a function of sea
level rise. Patches were identified by a set of
macroecological criteria, such as area, habitat
suitability, and neighboring distance, as a function of
the maximum dispersal. The distribution of the
predicted patch size was Korcak's law, whose
exponent is half of the fractal dimension of the
patches. We validated the model by predicting the
observed patch-size distribution and patch patterns
from 2002 to 2010 where data were available. We also
investigated the perimeter-size relationship for
estimating the fractal dimension of the patches at a
higher level of complexity because of the calculation
of the perimeter. The fractal dimension provided by
the perimeter-size relationship provided a median
estimate between the values derived from Korcak's
law and the box-counting distribution. Korcak's law
provided the most optimistic scenario of
fragmentation in which the probability of finding
large patches was the highest, while the box-counting
provided the most pessimistic scenario. Hence, the
perimeter-area relationship is suggested as the best
method to calculate the fractal dimension of the
mosaic of habitat patches.
* The robustness of the Pareto-L6vy distribution of the
patch size was verified for predictions of patches
from 2006 to 2100. Thus, the scale-invariance of the
patch patterns holds in time despite the strong
influence of sea level rise. This may be related to a
sort of simulated "biological resilience" of species to
the external changes (Folke et al. 2004) by assuming
invariant habitat area and dispersal requirement.
Scale-free habitat patterns have proven to be the
most resilient to external stressors in previous studies
(Kefi et al. 2011). Thus, the shape of the patch-size


probability and the fractal dimension when this
probability is a power law can be useful indicators to
estimate the "degree of stress" of coastal ecosystems.
Further research is anticipated to understand when
and how the patch-size probability deviates from a
Pareto-L6vy behavior. The fragmentation, which is
proportional to the fractal dimension of the
habitat-specific coastline, varied considerably over
time and in particular for the Piping Plover. However,
the risk of extirpation in 2100 for SNPL, PIPL, and
REKN was not high with respect to 2006. We note
that the comparison between final and initial years'
risk should not be the only comparison in evaluating
the risk of decline of a species. The overall trend of
the fractal dimension in the modeled period has to be
evaluated as well.
* A scaling relationship was found between the fractal
dimensions of the patches and of the habitat-specific
coastline. The scaling exponent of this relationship
appears to be species-independent for the shorebirds
considered. Further research is needed to explore the
conditions of universality (species- and ecosystem-
wise) of this relationship, which may be related to the
species considered. The fluctuation in the fractal
dimension of the coastline can be assumed to be a
valuable ecological indicator for assessing variation in
patch patterns of breeding and wintering shorebirds.
* We demonstrated that habitat loss, fragmentation,
and connectivity are three separate concepts.
Although these variables are closely linked to each
other, their causality is not trivial. For the shorebirds
studied, the predicted fragmentation was coupled
with habitat loss while the connectivity increased. The
fact that the patches, even if smaller, were connected
is an extremely positive factor that ensures dispersal
and gene flow; thus, the connectivity of patches
enhances the survivability of shorebirds. Birth, death,
and dispersal processes of a species can overcome the
habitat-loss effect and a decrease in the average size
of patches. Yet, a lower metapopulation risk of
extirpation exists if interpatch migration is allowed
(Kindvall and Petersson 2000). However, a decrease
in the average patch size can potentially increase
intra-species competition for foraging (Ritchie 1998)
and decrease carrying capacity. A possible optimal
ecogeomorphological state of the coastal ecosystem
may be characterized by the smallest fractal
dimension of the coastline that maximizes the
compactness of the suitable patches. This
configuration also minimizes the fractal dimension of
the patches. The highest entropy of this configuration
may translate into the smallest energy expenditure of
the species that inhabit the habitat, for example, for
foraging and breeding activities. The entropy of


Page 14 of 17








Convertino etal. Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


geomorphological landforms (Nieves et al. 2010) may,
in fact, be highly correlated with the scale-invariance
of ecological patterns such as species-patch patterns.


Endnotes
submitted to Ecological Processes Special Issue "Wet-
lands In a Complex World", Guest Editor: Dr. Matteo
Convertino



Additional files


Additional file 1: Additional Methods, Additional Results and
Discussion, Additional Tables, and Additional Figure.
Additional file 2: Video S1. Predicted land cover by SLAMM from 2006 to
2100.


Abbreviations
SNPL: Snowy Plover; PIPL: Piping Plover; REKN: Red Knot; TER: threatened,
endangered, and at risk; SLAMM: Sea Level Affecting Marshes Model; SLR sea
level rise; Df: fractal dimension of the coastline (from box-counting); Db: fractal
dimension of the breeding and wintering occurrences (from box-counting);
DK: fractal dimension of the patches (from Korcak's law); Dc: fractal dimension
of the patches (from perimeter-size relationship); S: patch-size; p: patch
perimeter; P(hs): habitat suitability score; SI: suitability index; Sp: minimum
population patch-size; Sb/,: minimum breeding/wintering patch-size; hr:
home-range; hrd: home-range distance; dl: maximum dispersal length; MLE:
maximum likelihood estimation.

Competing interests
The authors declare that they have no competing interests.


Author's contributions
MC designed the study, managed and analyzed the data, wrote the model
(box-counting and patch delineation model), developed the theory, and
wrote the manuscript. AB assisted in making the calculations and analysis, and
helped in writing the manuscript. GAK and RMC participated in the habitat
suitability modeling framework and reviewed the manuscript. IL supervised
the whole work, and reviewed the manuscript by providing a practical angle
to this research for effective environmental management. All authors read and
approved the final manuscript.


Authors' information
MC is Research Scientist at the University of Florida, Gainesville, and a
Contractor of the Engineering Research and Development Center of the US
Army Corps of Engineers at the Risk and Decision Science Team. AB is currently
a financial analyst at Frontier Airlines. AB got his B.Sc and M.Sc. from MIT, Civil
and Environmental Engineering program. AB performed his research
internship at the Risk and Decision Science Team in the summer of 2011. GAK
and RMC are Associate and Professor at the University of Florida, Gainesville,
respectively. IL is team leader of the Risk and Decision Science Team of the
Engineering Research and Development Center of the US Army Corps of
Engineers.


Acknowledgements
This research was supported by the US Department of Defense, through the
Strategic Environmental Research and Development Program (SERDP), Project
SI-1699. M.C. acknowledges the funding of project "Decision and Risk Analysis
Applications Environmental Assessment and Supply Chain Risks" for his
research at the Risk and Decision Science Team. The computational resources
of the University of Florida High-Performance Computing Center (http://hpc.
ufl.edu) are kindly acknowledged. The authors cordially thank Dr. RA Fisher
(Engineering Research and Development Center of the US Army Corps of
Engineers) and the Eglin Air Force Base personnel for their help in obtaining


the data and for the useful information about the breeding information of
SNPL. Tyndall Air Force Base and Florida Wildlife Commission are also
gratefully acknowledged for the assistance with the data. We thank M.L.
Chu-Agor (currently at the Center of Environmental Sciences, Department of
Biology and Earth and Atmospheric Sciences, Saint Louis University, St. Louis,
MO) for her computational effort with SLAMM at the University of Florida.
Permission was granted by the USACE Chief of Engineers to publish this
material. The views and opinions expressed in this paper are those of the
individual authors and not those of the US Army or other sponsor
organizations.

Author details
SDepartment of Agricultural and Biological Engineering- IFAS, University of
Florida, Gainesville, FL, USA. Contractor at the Risk and Decision Science
Team, Environmental Laboratory, Engineer Research and Development
Center, US Army Corps of Engineers, Concord, MA, USA.3 Florida Climate
Institute, UF-FSU, c/o Frazier Rogers Hall, Gainesville, FL USA. 4Department of
Civil and Environmental Engineering, Massachusetts Institute ofTechnology,
Cambridge, MA, USA. Department of Earth, Atmospheric, and Planetary
Science, Massachusetts Institute ofTechnology, Cambridge, MA, USA. "Risk
and Decision Science Team, Environmental Laboratory, Engineer Research and
Development Center, US Army Corps of Engineers, Concord, MA, USA.
7Department of Engineering and Public Policy, Carnegie Mellon University,
Pittsburgh, PA, USA.

Received: 11 June 2012 Accepted: 11 September 2012
Published: 30 October 2012

References
Aiello-Lammens ME, Chu-Agor ML, Convertino M, Fischer RA, Linkov I, Resit
Akcakaya H (2011) The impact of sea level rise on snowy plovers in Florida:
integrating geomorphological, habitat, and metapopulation models.
Global Change Biol 17:3644-3654. doi:1 0.1111/j.1365-2486.2011.02497.x
Airoldi L (2003) Effects of patch shape in intertidal algal mosaics: roles of area,
perimeter and distance from edge. Mar Biol 143: 639-650
Alliance FS (2010) Florida Panhandle Shorebird Working Group. Tech. rep
FWC/Audubon Florida/USFWS. http://www.flshorebirdalliance.org/
resources-pages/maps.html
Audubon (2006) America's top ten most endangered birds. Tech. rep. National
Audubon Society, New york
Bak P (1999) How nature works: the science of self organized criticality.
Springer-Verlag, Berlin
Baldassarri A, Sapoval B, Felix S (2012) A numerical retro-action model relates
rocky coast erosion to percolation theory. Arxiv. iii .. 1 1 1202.
4286vl.pdf
BanavarJR, Colaiori F, Flammini A, Maritan A, Rinaldo A (2001) Scaling,
optimality and landscape evolution. J Stat Phys 104:1-33
Batty M, Longley, P (eds) (1994) Fractal cities. Academic Press, San Diego
Bonabeau E, Dagorn L, Freon P (1999) Scaling in animal group-size
distributions. Proc Nat Acad Sci 96: 4472-4477. doi:10.1073/pnas.96.8.4472
Buldyrev S, Dokholyan N, Erramilli S, Hong M, Kim J, Malescio G, Stanley H
(2003) Hierarchy in social organization. Phys Stat Mech Appl 330(3-4):
653-659. doi:10.1016/j.physa.2003.09.041
Burney C (2009) Florida beach-nesting bird report, 2005-2008. Tech. rep.
Florida Fish and Wildlife Conservation Commission, Tallahassee. http://
ww i, l. i ,, i, ..- ,i ii ,)05-2008_FWCBNBReport.pdf
C-CAP (2009) Coastal change analysis program regional land cover. Tech. rep.
NOAA, Washington DC. http://www.csc.noaa.gov/digitalcoast/data/
ccapregional/
Chu-Agor M, Muhoz-Carpena R, Kiker G, Emanuelsson A, Linkov I (2011)
Exploring sea level rise vulnerability of coastal habitats using global
sensitivity and uncertainty analysis. Environ Model Software 26(5): 593-604
Clough J (2006) Application of SLAMM 4.1 to nine sites in Florida. Warren
Pinnacle Consulting, Waitsfield, VT. http://warrenpinnacle.com/prof/
SLAMM/NWF.SLAMMFLORIDA.2-16- 2006.doc
Clough JS (2010) The Sea Level Affecting Marshes Model. Tech. rep. Warren
Pinnacle Consulting, Waitsfield, VT. http://warrenpinnacle.com/prof/
SLAMM6/SLAMM6_Technical.Documentation.pdf
Convertino M, Kiker G, Chu-Agor M, Muhoz-Carpena R, Martinez C,
Aiello-Lammens M, Akcakaya H, Fisher R, Linkov I (2010) Integrated


Page 15 of 17








Convertino et ol Ecological Processes 2012, 1:9
http://www.ecologicalprocesses.com/content/1/1/9


modeling to mitigate climate change risk due to sea-level rise of imperiled
shorebirds on Florida coastal military installations. In: Linkov I, BridgesT
(eds) Climate change: global change and local adaptation. Springer,
Dordrecht
Convertino M, Donoghue J, Chu-Agor M, Kiker G, Munoz-Carpena R, Fischer R,
Linkov I (201 la) Anthropogenic renourishment feedback on shorebirds: a
multispecies Bayesian perspective. Ecol Eng 37: 1184-1194.
doi:10.1016/j.ecoleng.2011.02.019
Convertino M, Kiker G, Munoz-Carpena R, Chu-Agor M, Fischer R, Linkov I
(2011 b) Scale- and resolution-invariance of suitable geographic range for
shorebird metapopulations. Ecol Complexity 8(4): 364-376.
doi:10.1016/j.ecocom.2011.07.007
Convertino M, Elsner J, Munoz-Carpena R, Kiker G, Fisher R, Linkov I (2011 c) Do
tropical cyclones shape shorebird habitat patterns? Biogeoclimatology of
Snowy Plovers in Florida. PLoS ONE 6(1): pl
Convertino M, Simini F, Catani F, Kiker G (2012) From river basins to elephants
to bacteria colonies: aggregate-size spectrum of animate and inanimate
species. PLoS ONE (in press)
del Barrio G, Harrison P, Berry P, Butt N, Sanjuan M, Pearson R, Dawson T (2006)
Integrating multiple modelling approaches to predict the potential
impacts of climate change on species' distributions in contrasting regions:
comparison and implications for policy. Environ Sci Policy 9(2): 129-147.
doi:10.1016/j.envsci.2005.11.005
Ehrlich PR, Levin SA (2005) The evolution of norms. PLoS Biol 3(6): el94.
doi:10.1371/journal.pbio.0030194
Elliott Smith E, Haig S (2004) Piping plover (charadrius melodus). In: Poole A
(ed) The Birds of North America Online. Cornell Lab of Ornithology, Ithaca,
doi:10.2173/bna.2
F-DEP, State Geological Map FDEP-GEO (2001).Tech. rep., Florida Department
of Environmental Protection (data from, FGDL, University of Florida
GEOPLAN Center). http://www.dep.state.fl.us/geology/gisdatamaps/
state-geo-map.htm
Fallon F (2005) Petition to list the red knot (caladris canutus rufa) as endangered
and request for emergency listing under the endangered species act. Tech.
rep. Delaware Riverkeeper Network, American Littoral Society, Delmarva
Ornithological Society, Delaware Chapter of the Sierra Club, New Jersey
Audubon Society. http://w w i 11i. I I,. ,1 i i 1i i
Folke C, Carpenter S, Walker B, Scheffer M, Elmqvist T, Gunderson L, Holling C
(2004) Regime shifts, resilience and biodiversity in ecosystem
management. Annu Rev Ecol Evol Syst 35: 557-581
FWC (2010) Florida beach-nesting birds website. Tech. rep. Florida Fish and
Wildlife Conservation Commission. http://legacy.myfwc.com/bnb/
Harrington B (2001) Red knot (calidris canutus). Poole A (ed) The Birds of North
America Online. Cornell Lab of Ornithology, Ithaca, doi:10.2173/bna.563
Helzer C, Jelinski D (1999) Ecol App 9:1448-1458
Himes J, Douglass N, Pruner R, Croft A, Seckinger E (2006) Status and
Distribution of Snowy Plover in Florida, 2006. Tech. rep. Florida Wildlife
Conservation Commission. http://ww 11 1. I 1v I .11 ... I 1 '
Himes_Douglass-2006_SNPL_Report.pdf
Humphries N, et al. (2010) Environmental context explains Levy and Brownian
movement patterns of marine predators. Nature 465(7301): 1066-1069.
doi:10.1038/nature09116
Imre AR, Bogaert J (2004) The fractal dimension as a measure of the quality of
habitats. Acta Biotheoretica 52:41-56
Jovani R, Tella J (2007) Fractal bird nest distribution produces scale-free colony
sizes. Proc R Soc B 274: 2465-2469. doi:10.1098/rspb.2007.0527
Jovani R, Serrano D, Ursua E, Tella JL, Adler FR (2008) Truncated power laws
reveal a link between low-level behavioral processes and grouping
patterns in a colonial bird. PLoS ONE 3: 1992-+.
doi:10.1371/journal.pone.0001992
K6fi S, Rietkerk M, Alados CL, Pueyo Y, Papanastasis VP, Elaich A, de Ruiter PC
(2007) Spatial vegetation patterns and imminent desertification in
Mediterranean arid ecosystems. Nature 449:213-217.
doi:10.1038/nature06111
Kefi S, Rietkerk M, Roy M, Franc A, de Ruiter P, Pascual M (2011) Robust scaling
in ecosystems and the meltdown of patch size distributions before
extinction. Ecol Lett 14(1): 29-35. doi:1 0.1111 /j.1461 -0248.2010.01553.x
Kindvall O, Petersson A (2000) consequences of modelling interpatch migration
as a function of patch geometry when predicting metapopulation
extinction risk. Ecol Modell 129(1): 101 -109


Klemas V, Dobson J, Ferguson R, Haddad K (1993) A coastal land cover
classification system for the noaa coastwatch change analysis project. J
Coastal Res 9(3): 862-872
Korcak J (1940) Deux types fondamentaux de distribution staffstique. Bull Inst
Int Star 30:295-307
Lamonte K, Douglass N (2002) Status and Distribution of Snowy Plover in
Florida, 2002. Tech. rep. Florida Wildlife Conservation Commission.
http://ww v i. i i i .I... i i i Lamonte-Douglass-
2002.SNPLReport.pdf
Leyrer J, Spaans B, Camara M, Piersma T (2006) Small home ranges and high
site fidelity in red knots (calidris c. canutus) wintering on the banc dArguin,
mauritania. J Ornithology 147(2): 376-384. doi:10.1007/s10336-005-0030-8
Majka D, Jenness J, Beier P (2007) Corridordesigner: Arcgis tools for designing
and evaluating corridors. CorridorDesigner. http://corridordesign.org
Mandelbrot B (1967) How long is the coast of britain? Statistical self similarity
and fractional dimension. Science 156(3775).
doi:1 0.1126/science.l56.3775.636
Mandelbrot, B (1982) The Fractal Geometry of Nature. W.H. Freeman, New York
Masetti R, Fagherazzi S, Montanari A (2008) Application of a barrier island
translation model to the millennial-scale evolution of Sand Key, Florida.
Continental Shelf Res 28: 1116-1126. doi:10.1016/j.csr.2008.02.021
Morais PA, Oliveira EA, Araujo NAM, Herrmann HJ, Andrade JS (2011) Fractality
of eroded coastlines of correlated landscapes. Phys Rev E 84:016,102.
doi:10.1103/PhysRevE.84.016102
Muller S, Munoz-Carpena R, Kiker G (2010) Model relevance: frameworks for,
exploring the complexity-sensitivity-uncertainty trilemma. Linkov I, Bridges
T (eds) Climate Change: Global Change and Local Adaptation. NATO book,
Amsterdam
National Research Council, NAS (ed) (2009) A New Biology for the 21st
Century. The National Academies Press, Washington, DC
Nieves V, Wang J, Bras R, Wood E (2010) Maximum entropy distributions of
scale-invariant processes. Phys Rev Lett 105(11): 118,701+.
doi:1 0.1103/PhysRevLett.l 05.118701
Nikora V, Pearson C, Shankar U (1999) Scaling properties in landscape patterns:
New Zealand Experience. Landscape Ecol 14:17-33
Page G, Stenzel L, Warriner J, Paton P (2009) Snowy plover (charadrius
alexandrinus). In: Cornell I Lab of Ornithology, Poole A (eds) The Birds of
North America Online, vol. 5. doi:10.2173/bna.154, http://bna.birds.cornell.
edu/bna/species/154
Paola C, Leeder M (2011) Environmental dynamics: Simplicity versus
complexity. Nature 469: 38 39. doi:10.1038/469038a
Pascual M, Manojit R, Guichard F, Flierl G (2002) Cluster-size distributions:
signatures of self organization in spatial ecologies. Phil Trans R Soc Lond B
357: 657-666. doi:10.1098/rstb.2001.0983
Pascual M, Roy M, Laneri K (2011) Simple models for complex systems:
exploiting the relationship between local and global densities. Theor Ecol
4(2): 211-222
Paton P, Edwards T (1996) Factors affecting interannual movements of snowy
plovers. The Auk 113(3): 534-543. doi:10.2307/4088973
Phillips S, Anderson R, Schapire R (2006) Maximum entropy modeling of
species geographic distributions. Ecol Modell 190(3-4): 231-259.
doi:10.1016/j.ecolmodel.2005.03.026
Phillips SJ, Miroslav D (2008) Modeling of species distributions with maxent:
new extensions and a comprehensive evaluation. Ecography 31(2):
161-175
Pruner R (2010) Assessing habitat selection, reproductive performance, and
the affects of anthropogenic disturbance of the Snowy Plover along the
Florida Gulf coast. Master's thesis, University of Florida, Gainesville, USA
Reinhardt L, Jerolmack D, Cardinale BJ, Vanacker V, Wright J (2010) Dynamic
interactions of life and its landscape: feedbacks at the interface of
geomorphology and ecology. Earth Surf Processes Landforms 35: 78-101.
doi:1 0.1002/esp.1912
Rinaldo A, Dietrich WE, Rigon R, Vogel GK, Rodrlguez-Lturbe I (1995)
Geomorphological signatures of varying climate. Nature 374: 632-635.
doi:10.1038/374632a0
Ritchie M (1998) Scale-dependent foraging and patch choice in fractal
environments. Evolutionary Ecol 12:309-330
Scanlon TM, Caylor KK, Levin SA, Rodriguez-Iturbe I (2007) Positive feedbacks
promote power-law clustering of Kalahari vegetation. Nature 449:
209-212. doi:10.1038/natureO6060


Page 16of 17








Convertino etal. Ecological Processes 2012, 1:9 Page 17 of 17
http://www.ecologicalprocesses.com/content/1/1/9




Schneider D, Tella J, Scaling theory: application to marine ornithology (2002).
Ecosystems 5: 736-748. doi:10.1007/s10021 -002-0156-y
Schwimmer RA (2008) A temporal geometric analysis of eroding marsh
shorelines: can fractal dimensions be related to processes. J Coastal Res
1: 152-158
Seavey JR, Gilmer B, McGarigal KM (2010) Effect of sea level rise on piping
plover (charadrius melodus) breeding habitat. Biol Conserv.
doi:10.1016/j.biocon.2010.09.017
Sole R, Bascompte, J (Eds) (2006) Self organization in Complex Ecosystems.
Princeton University Press, Princeton, NJ
Stenzel L, Warriner J, Warriner J, Wilson K, Bidstrup F, Page G (1994)
Long-distance breeding dispersal of Snowy Plovers in western North
America. J Animal Ecol 63: 887-902
Stenzel L, Page G, Warriner J, Warriner J, George D, Eyster C, Ramer B, Neuman
K (2007) Survival and natal dispersal of juvenile snowy plovers (Charadrius
alexandrinus) in central coastal California. The Auk 124(3): 1023-1036
USFWS (2009) Pping Rover (Charaddus melodus) 5 Year Review Summary and
Evaluaton. Tech. rep, US Fish and Wildlife Service. http://www.fws.gov/northeast/
endangered/PDF/Piping_PloverJfive_yearreviewand_summary.pdf
USGS (2010) National Elevation Dataset. Tech. rep., United State Geological
Survey. http://ned.usgs.gov/
USGS-FWS (2009) Data from the 2006, International Piping Plover Census
(IPIPLC), Tech. rep., United State Geological Survey and USFWS. http://
wv Il ,i i 1.. i i ii i i ii I Smith-Haig-2006-PIPLReport.pdf
Warriner JS, Warriner JC, Page GW, Stenzel LE (1986) Mating system and
reproductive success of a small population of polygamous Snowy Plovers.
Wilson Bull 98: 15-37

doi:10.1186/2192-1709-1-9
Cite this article as: Convertino etaL Shorebird patches as fingerprints of
fractal coastline fluctuations due to climate change. Ecological Processes
2012 1:9,



























Submit your manuscript to a SpringerOpenG
journal and benefit from:

> Convenient online submission
> Rigorous peer review
> Immediate publication on acceptance
> Open access: articles freely available online
> High visibility within the field
> Retaining the copyright to your article


Submit your next manuscript at D springeropen.com




Full Text

PAGE 1

ShorebirdPatchesasFingerprintsofFractal 1 CoastlineFluctuationsduetoClimateChange: 2 AdditionalFiles 3 M.Convertino, 1 ; 2 ; 3 A.Bockelie, 2 ; 4 ; 5 G.A.Kiker 1 ; 3 R.Mu~noz-Carpena, 1 ; 3 I. 4 Linkov 2 ; 6 5 1 DepartmentofAgriculturalandBiologicalEngineering-IFAS,Universityof 6 Florida,Gainesville,FL,USA 7 2 RiskandDecisionScienceTeam,EnvironmentalLaboratory,Engineer 8 ResearchandDevelopmentCenter,USArmyCorpsofEngineers,Concord, 9 MA,USA 10 3 FloridaClimateInstitute,UF-FSU,c/oFrazierRogersHall,Gainesville,FL 11 4 DepartmentofCivilandEnvironmentalEngineering,Massachusetts 12 InstituteofTechnology,Cambridge,MA,USA 13 5 DepartmentofEarth,Atmospheric,andPlanetaryScience,Massachusetts 14 InstituteofTechnology,Cambridge,MA,USA 15 6 DepartmentofEngineeringandPublicPolicy,CarnegieMellonUniversity, 16 Pittsburgh,PA,USA 17 August17,2012 18 Keywords:landcoverchange,coastalwetlands,coastlinecomplexity,fractaldimension, 19 habitatsuitability,patches,sea-levelrise 20 submittedtoEcologicalProcessesSpecialIssueonWetlandsInaComplexWorld, 21 GuestEditor:Dr.MatteoConvertino 22

PAGE 2

AdditionalMethods 23 MaximumLikelihoodEstimationofthePatch-sizeProbability 24 TheMaximumLikelihoodEstimationMLEmethodemployedisfullydescribedinClauset 25 etal..Inthisstudy,Paretopower-law,truncatedPareto-Levytruncatedpower26 law,andexponentialdistributionsaretestedforpatch-size S whichistherandomvariable 27 ofourinterest.TheappropriateMLEequationforthedistributionof S isusedtoderivean 28 exponentwiththeinitial s min parametersettotheminimumvaluefoundinthedataset.A 29 besttdatasetisgeneratedwiththeestimatedparametersandaKolmogorov-SmirnovKS 30 testisusedtodeterminethegoodnessofttheKS-Dstatistic.TheKStestistheaccepted 31 testformeasuringdierencesbetweencontinuousdatasetsunbinneddatadistributionsthat 32 areafunctionofasinglevariable.Thisdierencemeasure,theKS-Dstatistic,isdenedas 33 themaximumvalueoftheabsolutedierencebetweentwocumulativedistributionfunctions. 34 TheDstatisticbetweentwodierentcumulativedistributionfunctions P N 1 s and P N 2 s 35 isdenedby D =max
PAGE 3

inthoseareasislargeduetothelowelevationandslope.Thestripedpatternsobservedin 55 theEastApalacheeBayandintheSouthWestareasaretheresultofthe SLAMM model 56 thatconsidersexclusivelythechangeoflandcoverfromacellwithhigherelevationtoa 57 cellwithlowerelevation,thusnotaectingtheneighboringcellsChu-Agoretal.,2010.In 58 FigureS2thepercentageofcellsofeachlandcoverclasswithrespecttothetotalhabitatis 59 plottedintime.Theglobalsea-levelriseof2mproduceslocalsea-levelriseeectsbecause 60 ofthedierentparametersassignedtoeachofthesevenzonesalongtheFloridacoastand 61 becauseofthenaturalheterogeneitiesofelevationandslopeofeachzone.Eachlandcover 62 class-thatrepresentalsowetlandtypesswamp,cypressswamp,salt-marsh,andmangrove 63 -respondstosea-levelrisedierentlyaccordingtothezonalgeomorphologicalproperties 64 alongthecoast.Adecreaseinareaintimeisnotobservedforallthelandcoverclasses. 65 Thesmoothdeclineoftheestuarinebeachandthemildincreaseoftheoceanbeach 66 reectthequitesmoothtrendofthefractaldimensionofthepatches D K forSNPLand 67 REKNintime.Asforthescrub/shrubandsaltmarshclasses,theirroughandabrupt 68 variationsintimeareobservedalsoforthefractaldimensionofthepatchesforthePIPL 69 Figure5,bwhichoccupiesalsothesehabitats.Itappearsevidentthedirectassociation 70 betweengeomorphologicalprocessessimpliedaslandcoverchangeandbiologicalprocesses 71 describedaspatches.Ahugelossoftheswampisobserved,strongvariationsofthe 72 tidalat,andalargeincreaseoftheestuarineopenwaterduetotheinundationand 73 consequentlossofotherhabitatareas. 74 FiguresS3,S4,andS5,showthesuitabilityindexSIforthebreedingseasonofSNPL, 75 andthewinteringseasonofPIPLandREKN,intheGulfcoastofFloridafrom2006to2100. 76 Thehabitatsuitabilityindexiscalculatedasspeciedinthemaintextofthemanuscript 77 seeMethods.InFiguresS3,S4,andS5thepatchesaredelineatedaccordingtheproposed 78 patch-delineationmodelandarereportedbeloweverySuitabilityIndexmap.Ahabitat 79 patchisdenedasaclusterofpixelsthataregoodenough,bigenough,andcloseenough 80 togethertosupportfeedingandbreedingbyaparticularspecies.Goodenough"meansthat 81 theyhavesucientresourcesforthesurvivabilityofthespecies.Bigenough"reectsthe 82 factthatthereistheneedtohaveenoughareatosupportatleastonebreedingunit,typically 83 consideredamatingpairofindividualswithoverlappinghomeranges.Closeenough"means 84 thatthepixelsmustbeclustered,ratherthandividedintoacheckerboardbytoomuch 85 interspersionwithpixelsofbadhabitat.Byaparticularspecies"weemphasizethefact 86 thatonespecies'breedingpatchmaybeanotherspecies'worsthabitatpatch.Fromthe 87 patternsofpatchesinFiguresS3,S4,andS5,thevariationofthepatchesisvisibleintime. 88 ThenumberofpatchesintheSouth-WestGulfcoastofFloridaismuchhigherforPIPLand 89 3

PAGE 4

REKNthanforSNPL,conrmingtheelddata.Thevariationofthepatchesislargerfor 90 PIPLthanforSNPLandREKN,especiallyintheFloridaPeninsulacoast. 91 TheconditionalprobabilityintimeofaSNPLbreedingoccurrenceandofaPIPLand 92 REKNadultoccurrenceinthewinterseasonisshowninFigureS6asafunctionofthe 93 predictedlandcoverFigureS1andthegeologylayersConvertinoetal.,2011b,a.The 94 conditionalprobabilitychangesaseachenvironmentalvariablelandcoverandgeologyis 95 varied,keepingallotherenvironmentalvariablesattheiraveragesamplevalue.Inother 96 words,theresponsecurvesofFigureS6showthemarginaleectofchangingexactlyone 97 variable,whereasthemodel MaxEnt maytakeadvantageofsetsofvariableschanging 98 together.TheresponsecurvesarederivedfromthehabitatsuitabilitymapsofFiguresS3, 99 S4,andS5.Fromtheresponsecurvesitispossibletonotethesimilarityofhabitatpreferences 100 betweentheSNPLinthebreedingseasonandthePIPL/REKNinthewinteringseasons. 101 SLAMM isthelandcoveraccordingthe SLAMM modelclassesandGEOisthegeology 102 layerConvertinoetal.,2011b,a.ThePipingPloverseemstooccupyalsoscrub/shrub 103 transitionalmarshandsaltmarshareasclass7,8,and9respectively.ThePIPLusesthe 104 oceanbeachmorethanSNPLandREKNdoclass12.PIPL,andREKN,ndmoresuitable 105 substrateswithmedium/nesandandsiltGEOclass4andshellysandandclayGEO 106 class8.SNPLinsteadarefoundonlyinhabitatscharacterizedbywhitemedium/nesand 107 andsiltclass4thatistypicaloftheFloridaGulfcoastPanhandleandPeninsula.Itisnot 108 observedasignicantvariationintimeoftheconditionalprobabilitiesasafunctionofthe 109 environmentalvariablesforthethreeTERspeciesconsidered.Theconditionalprobability 110 tothegeologylayershowsmorevariationintimethanforthelandcoverbecausethegeology 111 hasbeenconsideredunchangedintime,thusthepredictionislessreliable. 112 FigureS7showsthenumberofpatchesvs.theaveragesizeofpatchesforSNPL,PIPL, 113 andREKN.Thedotsarethebinnedaverageofalltherealizationforeachyearsimulated. 114 Theplotshowshowthenumberofpatches N p decreaseswiththesizeofpatches h s i that 115 increases.ThePIPLhasthesmallestpatchesonaverage,whileSNPLandREKNhave 116 similarpatchesofhigheraveragesizewithrespecttothePIPL. 117 References 118 Chu-Agor,M.L.,R.Mu~noz-Carpena,G.A.Kiker,A.Emanuelsson,andI.Linkov 119 Exploringsea-levelrisevulnerabilityofcoastalhabitatsusingglobalsensitivityandun120 certaintyanalysis," EnvironmentalModelingandSoftware ,accepted. 121 4

PAGE 5

Clauset,A.,C.R.Shalizi,andM.E.J.NewmanPower-lawdistributionsinempirical 122 data," SIAMRev. ,Vol.51,pp.661{703. 123 Convertino,M.,J.F.Donoghue,M.L.Chu-Agor,G.A.Kiker,R.Munoz-Carpena,R.A.Fis124 cher,andI.LinkovaAnthropogenicrenourishmentfeedbackonshorebirds:A 125 multispeciesBayesianperspective," EcologicalEngineering ,Vol.37,pp.1184{1194. 126 Convertino,M.,G.A.Kiker,R.Munoz-Carpena,M.L.Chu-Agor,R.A.Fischer,andI.Linkov 127 bScale-andresolution-invarianceofsuitablegeographicrangeforshorebird 128 metapopulations," EcologicalComplexity ,Vol.8,No.4,pp.364{376. 129 Humphries,N.E.,N.Queiroz,J.R.M.Dyer,N.G.Pade,M.K.Musyl,K.M.Schaefer,D.W. 130 Fuller,J.M.Brunnschweiler,T.K.Doyle,J.D.R.Houghton,G.C.Hays,C.S.Jones,L.R. 131 Noble,V.J.Wearmouth,E.J.Southall,andD.W.SimsEnvironmentalcontext 132 explainsLevyandBrownianmovementpatternsofmarinepredators," Nature ,Vol.465, 133 No.7301,pp.1066{1069. 134 5

PAGE 6

AdditionalTableCaptions 135 TableS1. Land-covermodelparameters. SLAMM modelparametersforthe7distinct 136 geomorphologicalareas:Pensacola-Eglin,Tyndall,EastApalacheeBay,Big-Bend 137 ,TampaBay,Ft.Meyers,andNorthEverglades.Theparametersthatare 138 dierentforeachregionareevidencedingrey. 139 140 TableS2. Modelresults.Fractaldimensionofthepatches D K ,numberofpatches N p 141 meanpatchsize h s i ,meannumberofconnectedpatches h c i thatisafunctionoftheesti142 matedmaximumdispersaldistance d l ,fortheTER-sspeciesconsideredintheyears2006, 143 2020,2040,2060,2080,and2100. 144 145 TableS3. Comparisonamongfractaldimensionsderivedfromdierentmethodsforthe 146 SNPLoccurrences,andfractaldimensionofthecoastline. D b D c ,and D K arethefractal 147 dimensionderivedfromthebox-counting,theperimeter-arearelationship,andtheKorcak's 148 law.Thepatchdelineationmodelestimates D K isvalidatedintheperiod2006-2010against 149 theobservations D b .Therelationshipbetween D f and D b istheevidenceofacorrelation 150 betweengeomorphologicalandecologicalprocesses. 151 152

PAGE 7

AdditionalFigureCaptions 153 FigureS1. Simulatedlandcoverpatterns.Predictedlandcoverintimerepresentedas 154 SLAMM classesfortheyears2006,2020,2040,2060,2080,and2100.Thedomainhasbeen 155 dividedintosevenareasbecausedierentsea-levelrisetrendandtidaldynamics:Pensacola156 Eglin,Tyndall,EastApalacheeBay,Big-Bend,TampaBay,Ft.Meyers 157 ,andNorthEverglades. 158 159 FigureS2. Predictedchangeoflandcoverintime.Variationintimeofthelandcover 160 classesfortheGulfcoastregionmodeled.afavoritehabitatforSNPL,PIPL,REKN 161 estuarinebeachandoceanbeach.blandcoverclassesfavorableforPIPL 162 scrub/shrub,saltmarsh,andmangrove.cotherclassesundevelopeddryland 163 ,swamp,cypressswamp,inlandfreshmarsh,tidalat,andestuarineopen 164 water.Manyclassesofthelandcoverarecoastalwetlandtypessalt-marsh,mangrove, 165 swamp,andcypressswamp.Theswampisthetypeofcostalwetlandsthatisexperiencing 166 thelargestdecayinarea. 167 168 FigureS3. SimulatedSNPLsuitabilityindexpatterns.SuitabilityIndexintimefor 169 theSNPLderivedfromthehabitatsuitabilitymodelHSM MaxEnt fortheyears2006, 170 2020,2040,2060,2080,and2100.Thesuitablepatchesdeterminedbythepatch-delineation 171 algorithmarerepresented. 172 173 FigureS4. SimulatedPIPLsuitabilityindexpatterns.SuitabilityIndexintimeand 174 predictedpatchesforthePIPLderivedfromtheHSM MaxEnt andthepatchmodelfor 175 theyears2006,2020,2040,2060,2080,and2100. 176 177 FigureS5. SimulatedREKNsuitabilityindexpatterns.SuitabilityIndexintimeand 178 predictedpatchesfortheREKNderivedfromtheHSM MaxEnt andthepatchmodelfor 179 theyears2006,2020,2040,2060,2080,and2100. 180 181 FigureS6. MaxEnt responsecurvesintime.Habitatsuitabilityforbreedingandwin182 teringforSNPL,PIPLandREKNrespectively,asafunctionofthelandcoveras SLAMM 183 classesandgeologyclassescalculatedby MaxEnt fortheyears2006,2020,2040,2060, 184 2080,and2100.TheSLAMMclassesarereportedinFigureS2.ThegeologyclassesGEO 185 are:clayeysand,gravelandcoarsesand,limestone,medium/nesandandsilt 186 ,sandyclayandclay,shellbeds,dolomite,shellysandandclay,lime187 7

PAGE 8

stone/dolomite,peat,andwater. 188 189 FigureS7. Averagenumberofpatchesvs.averagesize. N p isthenumberofpatches, 190 and h s i istheaveragesize.Thedotsarethebinnedaverageoftherealizationsforallthe 191 yearsmodeled. 192 193 8

PAGE 9

AdditionalVideo 194 VideoS1. Predictedlandcoverby SLAMM from2006to2100. 195 196 9

PAGE 10

TableS1: 1234567 Directionofshore S S S W W W W HistoricTrendmm/yr 2 : 1 0 : 75 1 : 38 1 : 8 2 : 36 2 : 4 2 : 02 MeanTidalLevel-NAVD88m 0000000 GreatDiurnalTideRangem 0 : 383 0 : 409 0 : 492 1 : 158 0 : 688 0 : 401 0 : 874 SaltElevationm 0 : 5745 0 : 6135 0 : 738 1 : 737 1 : 032 0 : 6015 1 : 311 MarshErosionhorz.m/yr 2222222 SwampErosionhorz.m/yr 1111111 TidalFlatErosionhorz.m/yr 0 : 2 0 : 2 0 : 3 0 : 2 0 : 2 0 : 2 0 : 2 RegionalFloodMarshAccretionmm/yr 5 : 6 5 : 6 5 : 6 2 : 1 2 : 1 2 : 5 2 : 5 IrregularFloodMarshAccretionmm/yr 3 : 753 : 753 : 763 : 753 : 753 : 753 : 75 TidalFreshMarshAccretionmm/yr 4 4 4 : 2 4 4 4 4 BeachSedimentationRatemm/yr 0 : 30 : 30 : 30 : 30 : 30 : 30 : 3 FrequencyOverwashyr 2222222 MaxWidthOverwashm 500500500500500500500 BeachtoOceanOverwashm 30303030303030 DrylandtoBeachOverwashm 30303030303030 EstuarytoBeachOverwashm 60606060606060 MarshPct.LossOverwash% 50505050505050 MangrovePct.LossOverwash% 25252525252525 RegionalFloodMaxAccretion.mm/yr 9 9 9 4 : 5 4 : 5 5 : 7 5 : 7 Reg.FloodMinAccretionmm/yr 4 : 7 4 : 7 4 : 7 1 : 4 1 : 4 1 : 4 1 : 4 Reg.FloodDiusionEectMaxm 0000000 Reg.FloodDiusionEectMin1111111 Reg.FloodSalinityTurbulenceMaxppt 0000000 Reg.FloodTurbulenceMaxZoneppt 0000000 Reg.FloodSalinityNonTurb.Max1111111 10

PAGE 11

TableS2: 200620202040206020802100 SnowyPlover D K 1 : 471 : 5761 : 5821 : 541 : 5731 : 55 N p 219182206226304312 h s i 10 : 379 : 388 : 488 : 327 : 597 : 92 h c i 577698 PipingPlover D K 1 : 71 : 641 : 6461 : 3561 : 5951 : 733 N p 369550357403996692 h s i km 2 6 : 487 : 036 : 336 : 3111 : 125 : 81 h c i km 12101191020 RedKnot D K 1 : 421 : 4521 : 4641 : 4691 : 6171 : 56 N p 189220177199272320 h s i km 2 11 : 30109 : 899 : 527 : 597 : 90 h c i km 111313151818 11

PAGE 12

TableS3: 200220032004200520062007200820092010 D f 1 : 261 : 281 : 371 : 351 : 291 : 291 : 301 : 321 : 33 SnowyPlover D b 1 : 631 : 621 : 751 : 741 : 631 : 641 : 661 : 681 : 70 D c )-2422()-2422()-2423()]TJ/F8 9.9626 Tf 122.541 0 Td [(1 : 521 : 531 : 581 : 611 : 65 D K )-2422()-2422()-2423()]TJ/F8 9.9626 Tf 122.541 0 Td [(1 : 471 : 491 : 521 : 541 : 58 12

PAGE 13

FigureS1:

PAGE 14

FigureS2:

PAGE 15

FigureS3:

PAGE 16

FigureS4:

PAGE 17

FigureS5:

PAGE 18

FigureS6:

PAGE 19

FigureS7:



PAGE 1

Convertino etal.EcologicalProcesses 2012, 1 :9 http://www.ecologicalprocesses.com/content/1/1/9 RESEARCHOpenAccessShorebirdpatchesas“ngerprintsoffractal coastline”uctuationsduetoclimatechangeMatteoConvertino1,2,3*,AdamBockelie2,4,5,GregoryAKiker1,3,RafaelMu noz-Carpena1,3andIgorLinkov6,7 AbstractIntroduction: TheFloridacoastisoneofthemostspecies-richecosystemsintheworld.Thispaperfocusesonthe sensitivityofthehabitatofthreatenedandendangeredshorebirdstosealevelriseinducedbyclimatechange,andon therelationshipofthehabitatwiththecoastlineevolution.WeconsidertheresidentSnowyPlover( Charadrius alexandrinusnivosus ),andthemigrantPipingPlover( Charadriusmelodus )andRedKnot( Calidriscanutus )alongthe GulfCoastofMexicoinFlorida. Methods: Weanalyzeandmodelthecoupleddynamicsofhabitatpatchesoftheseimperiledshorebirdsandofthe shorelinegeomorphologydictatedbylandcoverchangewithconsiderationofthecoastalwetlands.Thelandcoveris modeledfrom2006to2100asafunctionoftheA1Bsealevelrisescenariorescaledto2m.Usingamaximum-entropy habitatsuitabilitymodelandasetofmacroecologicalcriteriawedelineatebreedingandwinteringpatchesforeach yearsimulated. Results: Evidenceofcoupledecogeomorphologicaldynamicswasfoundbyconsideringthefractaldimensionof shorebirdoccurrencepatternsandofthecoastline.Ascalingrelationshipbetweenthefractaldimensionsofthe speciespatchesandofthecoastlinewasdetected.Thepredictedpowerlawofthepatchsizeemergedfrom scale-freehabitatpatternsandwasvalidatedagainst9yearsofobservations.Wepredictanoverall16%lossofthe coastallandformsfrominundation.Despitethechangesinthecoastlinethatcausehabitatloss,fragmentation,and variationsofpatchconnectivity,shorebirdsself-organizebypreservingapower-lawdistributionofthepatchsizein time.Yet,theprobabilityof“ndinglargepatchesispredictedtobesmallerin2100thanin2006.ThePipingPlover showedthehighest”uctuationinthepatchfractaldimension;thus,itisthespeciesatgreatestriskofdecline. Conclusions: Weproposeaparsimoniousmodelingframeworktocapturemacroscaleecogeomorphological patternsofcoastalecosystems.Ourresultssuggestthepotentialuseofthefractaldimensionofacoastlineasa “ngerprintofclimaticchangeeectsonshoreline-dependentspecies.Thus,thefractaldimensionisapotentialmetric toaiddecision-makersinconservationinterventionsofspeciessubjectedtosealevelriseorotheranthropicstressors thataecttheircoastlinehabitat. Keywords: Landcoverchange,Coastalwetlands,Coastlinecomplexity,Fractaldimension,Habitatsuitability, Patches,Sealevelrise *Correspondence:mconvertino@u”.edu 1 DepartmentofAgriculturalandBiologicalEngineering-IFAS,Universityof Florida,Gainesville,FL,USA 2 ContractorattheRiskandDecisionScienceTeam,EnvironmentalLaboratory, EngineerResearchandDevelopmentCenter,USArmyCorpsofEngineers, Concord,MA,USA Fulllistofauthorinformationisavailableattheendofthearticle 2012Convertinoetal.;licenseeSpringer.ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommons AttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,andreproduction inanymedium,providedtheoriginalworkisproperlycited.

PAGE 2

Convertino etal.EcologicalProcesses 2012, 1 :9 Page2of17 http://www.ecologicalprocesses.com/content/1/1/9IntroductionFloridacoastline-dependentspeciesarecharacterizedby oneofthehighestextirpationrisksintheworldbecause ofsealevelriseandincreaseintropicalcycloneactivity(Convertinoetal.2010,2011c)duetoclimatechange. TheSnowyPlover( Charadriusalexandrinusnivosus ; SNPLhereafter)isaresidentialshorebirdofFlorida listedasthreatenedatthestatelevel.ThePipingPlover ( Charadriusmelodus ;PIPLhereafter)isfederallydesignatedasthreatened,anditmigratesmostlyfromthe NorthAtlanticcoastsoftheUSAandCanadatoFlorida whereitwintersfor3monthsonaverage(ElliottSmith andHaig2004).TheRedKnot( Calidriscanutus ;REKN hereafter)isdesignatedasthreatenedinNewJerseyand isfederallylistedasapotentialatriskŽspecies.REKN usestheFloridaGulfbeachesasstop-overareasforabout 3weeksduringitsmigrationbetweenSouthAmerica andNorthAmericasBigLakesregionandAtlanticcoast (Harrington2001).Thisisconsideredasthewintering periodoftheREKNinFlorida.Anunderstandingof thespatialdistributionofthesuitablehabitatpatches fortheseshorebirds,theircontrollingfactors,andhow thesefactorsareaectedbysealevelriseisfundamentallyimportantforadoptingecientconservation strategies.Anunderstandingoflinkagesbetweenthecoupledevolutionoflandformsandecologicalpatternsis acrucialtopicduetotheevidencethatthesepatterns aretightlylinked.Biocomplexityapproaches(Mandelbrot 1982;Rinaldoetal.1995;Banavaretal.2001;Pascual etal.2002;SchneiderandTella2002;Buldyrevetal. 2003;delBarrioetal.2006;Sol eandBascompte2006; Scanlonetal.2007),despitebeingaccusedofadopting simpli“edbiologicalmodels(PaolaandLeeder2011),are capableofreproducingmacroscalepatternsofcomplex phenomenaandofdevelopingindicators,suchasthe probabilityofthepatchsize(Mandelbrot1982;Bonabeau etal.1999;JovaniandTella2007;K e“etal.2007;Jovani etal.2008;Convertinoetal.2012),thatareusefulfor assessingecosystemhealth(Ke“etal.2011).Oneofecologysmaingoalsistodetectfromobservedpatterns, suchasspeciesoccurrencepatterns,theorganizational rulesofspeciesinstationaryandevolvingecosystems. Manytheorieshavebeenproposedtoexplaintheformationofclusteredpatternsofspeciesinnature.Conspeci“c attraction,environmentalheterogeneities,andfoodavailabilityhavebeenclaimed„aloneortogether„tobethe motivationfortheformationofhabitatpatchesinwhich individualsofaspeciescoexistincolonies.Anoptimal searchtheory,theso-calledL evy-”ightforaginghypothesis(orpredator-prey-foodresourcedynamics),predicts thatpredatorsshouldadoptsearchstrategiesknownas L evy”ightswherepreyissparseanddistributedunpredictably.However,Humphriesetal.(2010)showedthat Brownianmovementissucientlyecientforlocating abundantprey.Thistheoryexplainstheclusteredpatterns ofresourcesinlandscapesthatmaybedierentfromthe patternofspeciesoccurrence.NeithertheL evy-”ightforaginghypothesisnorBrownianmovementmodeladdress thelinkagesofbiotawithlandscapeformsandtheirevolution,whichis,inouropinion,oneofthemainmissing points. Thecolonysizeofseabirds(SchneiderandTella2002), colonialbirds(JovaniandTella2007),andmanyother otheranimals(Bonabeauetal.1999)hasbeenfound tofollowapower-lawdistribution.Analogousscale-free distributionshavebeendetectedforbacteriacolonies (Buldyrevetal.2003),forspeciesincomplexecosystems(Sol eandBascompte2006;Convertinoetal.2012), andalsoforman-madesystemssuchascities(Battyand Longley1994).Theubiquityofthepower-lawstructure fortheprobabilityofthepatchsizeinaggregation phenomenaofnaturalandhumansystemssuggests theexistenceofuniversalself-organizationprinciples (Pascualetal.2002;Sol eandBascompte2006).Thescalingexponentofthepower-lawdistributionoftheaggregatesizewasproventobethefractaldimensionofthepatternanalyzed(Mandelbrot1982;Convertinoetal.2012). ThewordaggregateŽisageneralwordforindicatingthe assemblageofindividualswithsimilaroridenticalfeatures inalandscape.Inthepresenceofapowerlawfortheprobabilitydistributionoftheaggregatesize,theoccurrence patternsarescale-free,indicatingthatthepatternsare invariantatdierentscalesofobservations(Convertino etal.2012).TheconceptoffractaldimensionwasintroducedbyMandelbrotanalyzingthecoastlineofBritain atdierentscales(Mandelbrot1967).Theworkdisseminatedtheuseoffractalanalysis“rstingeomorphology (Moraisetal.2011;Baldassarrietal.2012)andlaterto avarietyofsciencesfrombiologytoengineering(Bak 1999).Nonetheless,allthesetheories,models,andempirical“ndingshaverarelyconsideredanypotentialeectof sloworabruptchangeintheexogenousfactorsonthe heterogenoushabitatinwhichspecieslive.Onlyrecently itwasprovenquantitativelythatecosystemsexhibitvariationsintheprobabilitydistributionofthepatchsize duetoanthropicallyandnaturallydrivenchangesinthe environmentalvariables(Ke“etal.2011).Forexample, deserti“cationofwater-controlledecosystemsproducesa decreaseinthefractaldimensionofvegetationpatches,or inextremecases,ashiftfromthepowerlawtoexponentialdistributionofthepatchsize(K e“etal.2007;Scanlon etal.2007;Ke“etal.2011).Climatechangescenarios testedintemperate/continentalregionsdepictedanoveralldecreaseinthefractaldimensionofpatchesintimefor manydierenttaxa(Barrioetal.2006).Forcolonialbirds thevariationinthefractaldimensionofthepatcheswasclearlyrelatedtothe”uctuationsinthepopulationabundanceduetointerspeciescompetition(Jovanietal.2008).

PAGE 3

Convertino etal.EcologicalProcesses 2012, 1 :9 Page3of17 http://www.ecologicalprocesses.com/content/1/1/9Ingeomorphologythevariationinthefractaldimension wasusedasthesignatureofthepersistingclimateover landscapes.Forexample,theassociationbetweenlandscapeevolutionandclimatehasbeenassessedforriver basinecosystemsinRinaldoetal.(1995).However,none ofthepreviousstudieslinkedthefractaldimensionoftwo ecosystemspatternsintime(e.g.,ofgeomorphological andecologicalpatterns)resultingfromlinkedprocesses. Hereweverifyforthe“rsttime,tothebestofourknowledge,thatthefractalityofthecoastlineisclearlylinkedto thehabitatpatchesofshoreline-dependentbirdsintheir breedingandwinteringseasons. Wehypothesizethatsealevelrisemayincreasethe complexityofthecoastlineandthatsuchcomplexity determinesfragmentationofthehabitatofspecies.We assumescaleinvarianceofthepatches,whichisalso detectablebytheanalysisoftheshorebirdoccurrences. Weconsiderabreedingshorebird(SnowyPlover)and winteringshorebirds(PipingPloverandRedKnot)in Floridatoquantifythepotentialeectofsealevelriseon residentandmigrantspecies.FortheSnowyPloverthe nestingseasonisusuallyconsideredpartofthebreeding season;thus,ourmodelsinputconsiderstheSNPLbreedingandnestingoccurrencessimultaneously.Furthermore, observationsindicatethatnesting,breeding,andwinteringareasforSNPLfallwithinthesamerange(Convertino etal.2011a).WinteringoccurrencesofSNPLarethus consideredtogetherwithbreedingoccurrences. Anintegratedecogeomorphologicalmodelingapproach isadoptedtopredicttheviabilityfrom2006to2100 ofthreatened,endangered,andatrisk(TER)shorebirds (SNPL,PIPL,andREKN)alongtheGulfCoastofFlorida asafunctionoftheincreasingsealevelriseduetoclimate change.Werescaleto2mtheIntergovernmentalPanel onClimateChange(IPCCA1B)scenariodescribedin Chu-Agoretal.(2011)andmodeltheecosystemata120 mspatialresolution.Wepredictlandcoverchangewith theSeaLevelAectingMarshesModel[SLAMM(Clough 2010)]whichisageomorphologicalmodelatlow-medium levelofcomplexity.SLAMMconsiderscoastalwetland typessuchasswamp,cypressswamp,mangrove,andsalt marsh(Additional“le1:FigureS1).Thehabitatmodel predictsthehabitatsuitabilityforbreedingandwinteringthroughamaximumentropyprincipleapproach (MAXENT)(PhillipsandMiroslav2008)asafunction oftherecordedspeciesoccurrencesinthebreedingand winteringseason,thepredictedlandcover,andageologylayer.MAXENTisanecologicalmodelatlowlevel ofcomplexity.Thelandcoverandhabitatsimulations areproducedinAiello-Lammensetal.(2011).Finally, inthispaperapatch-delineationmodelisintroducedto predicttheyearlyhabitatpatchesforasetofbiologicalconstraintsimposedonthehabitatsuitabilitymaps. Weassumethestationarityofthehabitatpatternsat theyearscaleandabsenceofbiologicaladaptationof speciestoclimatechange.Thefractaldimensionofthe patchesisderivedbythreeindependentmethods:(i) box-countingfortheobservedoccurrences;(ii)probabilitydistributionofthepatchsize[Kor cakslawŽ(Korcak 1940;Mandelbrot1982)];and(iii)perimeter-arearelationshipforthepredictedpatches.Weassumethatthese threemethodsproduceverycloseestimatesofthefractaldimensionofthewholemosaicofpatchesasshownin Convertinoetal.(2012). Thepower-lawdistributionofthepatchsizeisveri“ed byalmostadecade(2002…2010)ofhistoricalobservations ofthespecies.Thus,thepatch-delineationmodelisvalidatedagainsttheseobservationsfrom2002to2010.The coupledecogeomorphologicalorganizationisshownby thecorrespondenceintimeofthefractaldimensionsof thehabitat-speci“ccoastlineandofthepredictedpatches. Thefractaldimensionofthehabitat-speci“ccoastline, alongwithhabitatlossandpopulationabundance,is demonstratedtogreatlyin”uencethenumberandsizeof thepatches,whicharerelatedtohabitatlossandpopulationabundance.Althoughthefragmentationofthe habitat(whichisproportionaltothefractaldimension ofthepatches)ispredictedto”uctuateconsiderablyin thiscentury,theriskofextirpationofthespeciesanalyzedisnotdrasticallyincreasedbecausetheconnectivity ofthepatchesispredictedtoincrease.ThePipingPlover isthespecieswiththelargest”uctuationinthenumber andsizeofpatches.Webelievetheresearchpresented inthispaperconstitutesacontributiontotheemerging“eldofbiogeosciences,whichexplorestheinterface betweenbiologyandthegeosciencesandattemptsto understandtheinterrelatedfunctionsoflandscapesand biologicalsystemsacrossmultiplespatialandtemporal scales.Weareawareoftheexistenceofmanyothercomplexecogeomorphologicalprocessesthatarenotincluded inourmodelingeort.However,parsimoniousmodels suchasthemodelpresentedherecancapturelarge-scale patternswhilebypassingsmall-scaledetails(Ehrlichand Levin2005;Pascualetal.2011).Thesemodelscanbe testedagainstothermorebiologicallyrealisticmodelsto fullyexplorethelinkagesamongvariousenvironmentalchanges,geomorphologicaldynamics,andbiodiversity patterns.Weanticipatethatfurtherresearchwillexplore thisissueofprocesscomplexityversusmodelcomplexity,modelrelevance,andmodeluncertainty,whichcanbe synthesizedasamodelingtrilemmaŽ(Mulleretal.2010). Thispaperisorganizedasfollows.TheMethodsŽ sectiondescribestheshorebirddataandthestudysiteand explainsthemodelsusedinthisstudyandthetheoretical characterizationofpatches.TheResultsanddiscussionŽ sectionreportsthemainresultswithabroaddiscussion of“guresandhowtheseresultsareinterpretedconsideringourassumptions.TheConclusionsŽsectionreports

PAGE 4

Convertino etal.EcologicalProcesses 2012, 1 :9 Page4of17 http://www.ecologicalprocesses.com/content/1/1/9themostimportantconclusions,implicationsformanagement,andfurtherresearcheorts.Additional“les1and2 areprovidedtosupportourmainresult.MethodsSitedescriptionandbiogeographicalvariablesThewhite“ne-sandbeachesoftheFloridacoastofthe GulfofMexicoconstitutethehabitatofthewholeFlorida SNPLpopulation.TheSNPLpopulationinFloridaisdistributedalongabout80%oftheFloridaPanhandleand alongabout20%oftheFloridaPeninsula(Lamonteand Douglass2002;Himesetal.2006;Burney2009;Pruner 2010)(Figure1a).TheFloridaPeninsulaandtheAtlantic coastsarethemainwinteringgroundsforthemigratoryPIPLandREKN,whichseemlessconstrainedthan theSNPLbythemineralogicalpropertiesofthebeach substratecapturedbythegeologylayer(Convertinoetal. 2010,2011b).Thelandcover,whichincludesmanywetlandtypesfromC-CAP(2009)isrepresentedinFigure S1oftheAdditional“le1,andthegeology(F-DEP 2001)characterizesthemineralogicalsubstrateofeach landcoverclass(Additional“le1:FigureS6)(Convertino etal.2011b).In2006thePIPLPanhandle-Peninsulaand Atlanticpopulationswere38and33%,respectively,of thetotalmigrantPIPLpopulationinFlorida.TheREKN Panhandle-PeninsulaandAtlanticpopulationswere55 and20%,respectively,ofthetotalmigrantpopulation inFlorida.TheInternationalPipingPloverCensusin 2006supportedthe“eldsamplingofSNPL,PIPL,and REKN(USGS-FWS2009;FWC2010;Alliance2010). The2006winteringoccurrencesinFloridaarethedata usedinthisstudyforPIPLandREKN.FortheSNPL, dataofbreedingandnestingoccurrencesarealsoavailablefrom2002to2010andareprovidedyearbyyear bytheFloridaWildlifeCommission.Theseoccurrences areusedtoverifytheassumptionofscale-invariance ofSNPLoccurrencepatternsovertimewiththeboxcounting.However,despitetheavailabilityofSNPLdata from2002to2010,weconstructthehabitatsuitability modelwiththe2006SNPLoccurrencesaloneinorder tobeconsistentwiththe2006NOAAlandcover(CCAP2009)andthe2006PIPLandREKNoccurrences. ThegeologyandtheelevationfromUSGS(USGS2010; Convertinoetal.2010,2011b)areusedinthehabitat suitabilitymodelandinthelandcovermodel,respectively(Aiello-Lammensetal.2011;Convertinoetal.2010, 2011b). WeconsiderPIPLandREKNinthesamegeographic domainwherethefullrangeoftheSNPLoccursinorder toperformasimultaneousinterspeciesassessmentof thehabitatuseandextirpationriskofthethreespecies (Figure1a).Thus,onlythePanhandle-Peninsularegion wasconsideredinthisstudy.TheSNPLisourmaininterestbecauseitsyear-roundpresenceintheFloridacoastal ecosystemmakesthisspeciespotentiallymorevulnerablethanPIPLandREKN.DispersalamongthePanhandleandPeninsulaSNPLpopulationshasbeenobserved butnotquanti“ed.PopulationsubdivisionoftheSNPL hasnotbeenobserved;thus,wecanadoptthesame habitatanddispersalcriteriaforthewholepopulation. Populationsubdivision,forexample,canbecausedby geographicbarriersordisturbances[e.g.,renourishment (Convertinoetal.2011a)]thatinterferewiththedispersal.Thereductionindispersalisreportedtoreducegene ”owandincreasegeneticdriftofindependentsubpopulationsinthelong-term.However,thisisnotthecasefor theSNPLpopulationinFloridadespitetheweakinterchangeofindividualsbetweenPanhandleandPeninsula (Aiello-Lammensetal.2011). HabitatareaanddispersaldataforSNPLaremostly fromAiello-Lammensetal.(2011)butalsofromPage etal.(2009),PatonsandEdwards(1996),Stenzeetal. (1994;2007),Warrineretal.(1986).Aiello-Lammenset al.2011synthesizedthebiologicaldataandthemetapopulationmodelingeortofthisresearchfortheSNPL. Informationisgatheredalsofrom“eldecologistsworkingonthisproject[i.e.,Dr.R.A.Fischer(Engineering ResearchandDevelopmentCenter,USArmyCorpsof Engineers)andMrs.A.Pruner(FloridaParkService)].For PIPL,habitatanddispersaldataarefromAudubon(2006), Seaveyetal.2010,andUSFWS(2009),andforREKN,data arefromFallon(2005)andLeyreretal.(2006).Foramore detaileddescriptionofthesiteunderstudywereferthe readertoConvertinoetal.(2011b).Box-countingalgorithmThecharacterizationoftheoccurrencepatternsofbreedingandwinteringoccurrencesandofthecoastlineis performedusingthebox-countingŽmethod.Forthe SNPLtheoccurrencepatternofnestingoccurrenceswas observedtobeaself-similarpattern(Convertinoetal. 2012);thus,thebox-countingmethodissuitabletopredict howthispatternchangeswiththescaleofanalysis.The box-countinganalysisconsistsofcalculating,forgridsof dierentbox-sidelengths,thenumberofboxesthatcontaintheobjectunderstudy.Adjacentboxesconstitutean approximationoftherealpatchesateachresolution.The algorithmcanbeappliedtobothpointandlinepatterns. Thebox-countingisperformedovereightordersofmagnitudeinalogarithmicscaleofthebox-sidelength,from lo ( 1 )= 565km(whichcorrespondstotheboxB1Žin Figure1),whichisapproximativelythewidthofFlorida, to lo ( 5000 )= 11.3 10Š 3km(Figure1).Weindicatewith lo ( i )thelengthoftheboxsideatresolutiono=iŽfrom i = 1,...,5000wheretheincrementfromoneresolution toanotheris11.3 10Š 2km,whichisslightlysmaller thantheaveragehomerangeoftheSNPL(Table1).The orderofmagnitudeisrelativetothescalesofanalysis

PAGE 5

Convertino etal.EcologicalProcesses 2012, 1 :9 Page5of17 http://www.ecologicalprocesses.com/content/1/1/9 0200400s r e t e M 0 0 107,90015,800s r e t e M 0 5 9 3059,000118,000s r e t e M 0 0 5 9 2B3 B4 B50140,000280,000r e t e M 0 0 0 0 7B1 B2 Peninsula PanhandleRK 2006 PP 2006 SP 2006beach salt-marsh otherd)S p dlb) c) a) Figure1 Box-countingalgorithm. ( a )Representationofthethebox-countingalgorithmappliedtothe2006occurrencesoftheSnowyPlover (SNPL),PipingPlover(PIPL),andRedKnot(REKN),foreightordersofmagnitude(inalogarithmicscale),whichcorrespondsto5000resolutionsof thebox-countinggrid.InthisexampleattheresolutionofboxB5thenumberofboxesinwhichthereisatleastoneoccurrenceis N ( B 5 ) = 6.( b ) Box-countingexampleappliedtothewholecoastline,tothehabitat-speci“ccoastline(e.g.,beach,salt-marsh),andtootherlandcoverclassesas in Convertinoetal.(2011b).Manycoastalwetlandtypesareincludedinthelandcover,suchasswamp,cypressswamp,mangrove,andsaltmarsh.The shadedgridcellsin(a)and(b)haveatleastonespeciesoccurrenceoracoastlinesegmentattherepresentedresolution.Twocoastline con“gurationsarepresented:the“rstforhighvaluesof Dfand DK( c ),thesecondforlowvaluesof Dfand DK( d ).Thepatchespresentedingreen areconnectedbecausetheirneighboringdistanceislowerthanthemaximumdispersallength dl.

PAGE 6

Convertino etal.EcologicalProcesses 2012, 1 :9 Page6of17 http://www.ecologicalprocesses.com/content/1/1/9Table1Macroecologicalparametersofthe patch-delineationmodel,andbiologicaldataestimated fromtheliterature SNPLPIPLREKN Modelparameters Sp(km2)0.030.040.06 Sb / w(km2)0.010.020.04 dl(km)81220 Data hr (km2)0.0162.2010 15 hrd (km)0.121.48 2.403.10 nd (km)0.720.961.44 m (g)38 4050 60180 200 Spand Sb / waretheminimumpopulationandbreeding/winteringarea, respectively. dlistheestimatedmaximumdispersallength. nd isthe neighborhooddistanceinthebreedingseasonforSnowyPlover(SNPL),andin thewinteringseasonforPipingPlover(PIPL)andRedKnot(REKN). hr and hrd are thehomerangeandthehomerangedistanceestimatedconsideringthe breedingregionsforSNPL,PIPL,andREKN. m istheaveragebodymass.Dataare takenfromBurney(2009),Himesetal.(2006),LamonteandDouglass(2002), Pruner(2010),Aiello-Lammensetal.(2011).Modelparametersareassumed consideringdataandcalibratingthepatch-delineationmodelontheobserved patchsizein2006derivedfromthebox-counting.(extent)investigatedbythebox-counting,whiletheresolutionisrelatedtothegridschosenforthebox-counting. Thenumber N ( l ) ofboxesofsize l neededtocoverthe patternofoccurrences(whichisgenerallyafractalset) followsapowerlaw, N ( l ) = N0lŠ D,(1) where D d ,and d isthedimensionofthespace(usually d = 1,2,3). D isalsoknownastheMinkowski-Bouligand dimension,Kolmogorovcapacity,orKolmogorovdimension,orsimplythebox-countingdimensionandisanestimateoftheHausdordimension(Mandelbrot1982).The fractaldimensionfor1 Š d objectsisassociatedwiththe Hurstexponent H suchthat D = 2 Š H (Mandelbrot1982; Bak1999).ThevaluesoftheHurstexponentvarybetween 0and1,withhighervaluesindicatingasmoothertrend, lessvolatility,andlessroughnessoftheanalyzedpattern (Mandelbrot1982).Weindicatethefractaldimensionof thebreedingandwinteringoccurrenceswith Dbandthe fractaldimensionofthecoastlinewith Dfderivedfrom box-countinganalysis.Bothfractaldimensionsaredeterminedbythebox-countingmethod.Thefractaldimensionofthecoastlineiscalculatedalsoforeachlandcover classthatisaspecies-speci“chabitatforthespeciesconsidered(Figure1b).Manylandcoverclassesarecoastal wetlandtypes(Additional“le1:FigureS1).Land-covermodelThelandcoverispredictedyearbyyearbyusingtheSea LevelAectingMarshesModel(SLAMM)(Clough2006; Chu-Agoretal.2011)startingfromtheyear2006to2100. ThesesimulationsareperformedinAiello-Lammensetal. (2011)andConvertinoetal.(2010)towhichwerefer thereaderformoredetails.Thedomainofthemodel isextendedinlandforabout10kmfromthecoastline (Convertinoetal.2011a,2011b)(blackregionalongthe coastinFigure1,boxB1).Weconsiderthepredictedinundationdistancein2100( 9km)forarangeof[1,2]msea levelrise(SLR)adding1kmtoconsidertheuncertainty intheestimationofthe”oodingdistance.Theinitial conditionisthe2006landcoverfromNOAA(Klemas etal.1993).TheNOAAlandcoverclassesarechanged intoSLAMMlandcoverclassesformodelingpurposes. SLAMMrequiresustogrouptheclassesoflandcoverinto modelclasses.TheconversionisreportedinConvertino etal.(2011b).TheSLAMMmodelalsorequirestheelevationandslopeasinputvariables.Themodeleddomain isdividedintosevenregions(Additional“le1:Figure S1)withdistincthistoricaltidalandSLRtrends.Each regionischaracterizedbyauniquesetofvaluesforthe 26inputparameters(Additional“le1:TableS1)relatedto tide,accretion,sedimentation,anderosionprocesses.The valueoftheparametersisderivedfromtheavailableliteratureandpreviouseortsofthisresearch(Chu-Agoretal. 2011).Inthiseortofmodelingthelandcover,wedonot consideranygeomorphologicalfeedbackbetweenlandformsandclimatechangethatisexpectedtooccurwith globalwarming.Allourassumptionsarethesameasthose inChu-Agoretal.(2011)andConvertinoetal.(2010). Alsowedonotconsideranypossiblebarrierislandshiftingbecausethatisreportedtooccuroveratimeperiod muchlongerthanourpredictions(Masettietal.2008).HabitatsuitabilitymodelTheemployedhabitatsuitabilitymodelisMAXENT(Phillipsetal.2006;PhillipsandMiroslav2008),whichis oneofthemostdiusedmodelsinspeciesdistribution modeling.MAXENTisamodelbasedontheprincipleof maximumentropythatpredictscontinuoushabitatsuitabilitymapsofpotentialspeciesoccurrenceunderaset ofselectedenvironmentalvariables.Theenvironmental variablesthatarenecessaryandsucientforcalculating thehabitatsuitabilityarethelandcovertranslatedinto SLAMMclasses(Chu-Agoretal.2011;Convertinoetal. 2011a,2011b)andtheUSGSgeologylayer(Convertino etal.2011a,2011b)ataresolutionof120m.Theresolution 120misthehome-rangedistanceoftheSNPL(Table1). Suchdistanceissucienttocapturenotonlythespatial variabilityofhabitatpreferencesofSNPL,butalsothatof PIPLandREKN,whosehome-rangedistancearemuch largerthanthatofSNPL.Thehabitatsuitabilityat-a-point (i.e.,foreachpixelofthemodeleddomain)canbeconsideredasaproxyto“ndSNPL,PIPL,andREKNinthe breedingandwinteringseason.Thepriorprobabilitiesof

PAGE 7

Convertino etal.EcologicalProcesses 2012, 1 :9 Page7of17 http://www.ecologicalprocesses.com/content/1/1/9occurrencearecalculatedinMAXENTusingtherecorded shorebirdoccurrencesconstrainedtotheenvironmental variables.TheoccurrencesarenestandbreedingoccurrencesforSNPL,andtheadultoccurrencesforPIPLand REKN.Thus,forPIPLandREKNthehabitatsuitabilityreferstothesuitabilityforwinteringasinConvertino etal.(2011a).NoabsencesarerequiredinMAXENT. Thentheposteriorprobabilitiesofoccurrencearebased onthepriorprobabilitiesgiventhechangeintheland covermodeledyearbyyearbythelandcovermodel.A regularizationparameterthatcontrolsthe“tofthepredictedsuitabilitytotherealoccurrencedataisassumed tobeequaltoone.Non-randomlyplacedpseudoabsences areusedtoimprovethepredictions,and25%oftheoccurrencesaretakenasatrainingsample(Convertinoetal. 2011a,2011b).Thepredictedhabitatsuitabilitymapsrepresenttheaverageofover30replicatesforeachyearto reducetheuncertaintyofthepredictions.Thehabitat suitabilityiscalculatedwith10,000randombackground points.Backgroundpointsareasubsetofpointsofthe domainoverwhichtheBayesianinferencebetweenthe recordedspeciesoccurrences,pseudoabsences,andenvironmentallayersisdetermined. Weassignabiologicalinterpretationtothepredicted habitatsuitabilityscore, P ( hs ) ,whichistheprobabilityata-pointof“ndingabreedingand/orawinteringground. Breedingandwinteringgroundsaresuitablesitesforthe SNPLasafunctionoftheseasonconsidered,andwinteringgroundsaresuitablesitesforthePIPLandREKN.We de“nethesuitabilityindex(SI)asametricfrom0to100 thatcapturesthequalityofthebreedingand/orwintering habitatforthespecies.ThehighertheSIthelargerthebiologicalspectrumoffunctionsperformedbythespeciesin thathabitat.Hence, P ( hs ) isalsoasurrogateofhabitatuse duringthebreedingandwinteringseasonsofthespecies considered.Infact,itislegitimatetoassumethathabitatuseincreaseswithhabitatquality.Everypixelofthe HSmapsisclassi“edinto“veSIcategories:SI=100[for 0.8 P ( hs ) 1]isconsideredthebesthabitatwiththe highestsurvivaland/orreproductivesuccess;SI=80[for 0.6 P ( hs )< 0.8]istypicallyassociatedwithsuccessful breedingand/orwintering;SI=60[for0.2 P ( hs )< 0.6] isassociatedwithconsistentuseforbreedingandwintering;SI=30[for0.2 < P ( hs ) ]isassociatedwithoccasional usefornon-breeding,feedingactivities,andwintering;all valueslessthanSI=30indicatehabitatavoidedbothfor breedingandwintering;andSI=0forcompletelyunsuitablehabitat.WereferthereadertoConvertinoetal. (2011a)foradditionaldetailsaboutMAXENTrunsforthe SNPL,PIPLandREKN.Patch-delineationmodelBelowwede“neacriterionfordelineatingbreedingand winteringpatchesforSNPL,PIPLandREKN,respectively. Apatchisde“nedwhenthefollowingcriteriasimultaneouslyhold: for SI 60 [ i.e.,for P ( hs ) 0.2 ] P(X)withinaneighborhooddistance nd dlMinimumpopulationpatchsize SpMinimumbreeding/winteringpatchsize Sb / w. (2) Thespecies-dependentvaluesforthethreeparametersrequiredforthepatchidenti“cationarereportedin Table1.ThevaluesofbiologicaldatainTable1areused onlytosupportthechoiceofmodelparameters.The modelparametersarecalibratedtoreproduceapatch-size distributionascloseaspossibletothebox-countingdistributionofoccurrencesin2006.Themodelwiththisset ofparameterswasvalidatedagainstthepatch-sizedistributionsfrom2002to2010estimatedbythebox-counting. Wede“nebreedingpatchasanarealargeenoughtoat leastoccasionallysupportasinglebreedingpairthrough courtshipandrearingofyoungtodispersalage(Majka etal.2007).Apopulationpatchisde“nedasanarealarge enoughtosupportbreedingfor10yearsormore,evenif thepatchisisolatedfrominteractionwithotherpopulationsofthespecies(Majkaetal.2007).Sincepopulationwidedataarelackingforthesebreedingandpopulation arearequirements,weassumedthatapopulationpatch isatleasttwotimeslargerthanabreedingpatch.For theSNPLthesepatchescontaincertainnestingpatches. Theminimumpopulationandbreeding/winteringpatch areasareestimatedfromtheliteratureavailableand byexpertknowledgeofthe“eldbiologistsinvolvedin thesamplingcampaignsperformedforthisstudy(see Burney2009;Himes2006;LamonteandDouglass2002; Pruner2010). Spand Sb / waretheminimumpopulation andbreeding/winteringarea,respectively,andareproportionaltotheestimatedhomerange.Theminimum breeding/winteringareaistheminimumareathatwill supportbreedingandwinteringactivityoftheshorebirds. Thehomerange hr andthehome-rangedistance hrd (the squarerootofthe hr )arevaluesestimatedconsideringthe breedingregionsforSNPL,PIPL,andREKN.Weassume that Spand Sb / wforPIPL,andREKNaremuchsmaller than hr becausetheyrefertothewinteringperiodofthese shorebirdsinFlorida.ForREKN, Spisalsoreduceddueto thehabitatlimitationandtheclosecoexistencewithSNPL inthesamehabitat.Patchesareconsideredconnected iftheirneighboringdistanceisequaltoorsmallerthan dl,whichisthemaximumdispersallength.Figure1c,d showsanexampleofpatchesthatareconnectedbecause theirreciprocaldistanceislowerthan dl.Theseplots alsorepresentourassumptionthatcoastlinecomplexity aectspatchdistribution.Theaverageneighborhooddistance nd istheaveragedispersalofthespecies. nd is

PAGE 8

Convertino etal.EcologicalProcesses 2012, 1 :9 Page8of17 http://www.ecologicalprocesses.com/content/1/1/9higherthan hrd fortheSNPLduetothehigherlocaldispersalabilityestimatedfromrecentsurveys(Himesetal. 2006;Pruner2010).ForPIPLandREKN, nd issmaller than hrd becausethereported hrd referstotheirbreedingrangeinnorthernstatesintheUSAandCanada.In thewinterseasonPIPLandREKNmigratetoFlorida, andtheirdispersaldistanceisobservedtobesmaller. Withintheneighborhooddistanceasubpopulationcan beassumedtobepanmictic.Apanmicticpopulationis oneinwhichallindividualsarepotentialpartners.Itis usuallyestimatedfromtheforagingdistanceofananimal species.Inamoreabstractwaytheneighborhooddistance istheglueofallthesuitablepatches.Inaparticlephysics analogy,itdescribestheBrownianmotionofindividuals withinalargerspeciesgroup.Thus,byusing dl,whichis themaximumdispersal,asacriterioninthemodel,foragingiscertaintobeconsideredwithinpatches.Ourmodel considersanupperestimateofthepatchsizeforallthe shorebirdsconsidered. m istheaveragebodymass,which isusedtodiscusssomeresults.WeassumethesamebiologicalparametersfortheSNPLPanhandleandPeninsula asinAiello-Lammensetal.(2011).ProbabilitydistributionofthepatchsizeTheprobabilityofexceedanceofthepatchsizeisknown inliteratureasKor cakslaw(Korcak1940;Nikoraetal. 1999),whichisexpressedby: P ( S s ) = cSŠ F s sc ,(3) where c isaconstant, F isahomogeneityfunctionthat dependsonacharacteristicsize sc,and = DK/ 2isthe scalingexponent(Korcak1940;Mandelbrot1982). DKis thefractaldimensionofthepatches.Theprobabilityof exceedanceexhibitsapower-lawbehavior.Theprobability distributionofthepatchsizeforthepredictedpatcheswas usedtovalidatethepatch-delineationmodelagainstthe box-countingestimatesontherealoccurrencesfrom2002 to2010.The“tofthepredicteddistributionofpatchesis performedusingaMaximumLikelihoodEstimationtechnique(MLE),whichisdescribedintheAdditional“le1.Perimeter-arearelationshipThescalingrelationshipbetweentheperimeter p andthe size S ofthepatches: p = kSDc/ 2,(4) determinesthefractaldimensionofthemosaicofpatches, whichconsidersthefractalityofthepatchedge.Here weindicatethefractaldimension Dc,whichisderived fromthesamepredictedpatchesoftheintroducedpatch model(seethePatch-delineationmodelŽsection)but alsoconsiderstheirperimeters.BecauseKor cakslaw (Korcak1940)considersonlythesizeofthepatches,the perimeter-areascalinglawhasbeenconsideredasamore precisetoolformeasuringthefractaldimension.Inliteraturetheratio p / S isadoptedtomeasurethequalityof thepatchesforpopulationsurvivability,thatis,thelikelihoodofsurvivinginasuitablepatch(HelzerandJelinski 1999;Airoldi2003;ImreandBogaert2004).Ingeneralthe highertheratio p / S thelesssuitablethepatchareaforthe species,andthehighertheratio p / S ,thehigherthefractal dimension Dc.ResultsanddiscussionTherelationshipbetweenthenumberofcellsoccupiedby shorebirdoccurrences, N ( l ) ,andthelengthofthesideof thebox, l ,ateachscaleofanalysisisshowninFigure2. PP 2006 coastline 2006N( l )100102104106108105106104103102101100-1.135 -1.85 =a)-1.53 =RK 2006l N( l )100102104106108105106104103102101100b)Db SP1.56 1.66 1.76D b SPSP 2002 SP 2003 SP 2004 SP 2005 SP 2006 SP 2007 SP 2008 SP 2009 SP 2010Df=Db PPDb RK1.25 1.325 1.40 ‘10 ‘02 ‘06 ‘08D f year‘04(m) Figure2 Box-countingscaling-lawintime. ( a )Power-law N ( l ) = N0lŠ Dbderivedfromthebox-countingalgorithmappliedto theoccurrencesofPIPL(blackdots)andREKN(green)in2006andto thewholeFloridaGulfcoastline.IntheinsettheschematizedFlorida coastlineisevaluatedatdierentboxsizes.( b )Box-counting algorithmappliedtothe2002-2010occurrenceoftheSNPL.The fractaldimensionderivedfromtheanalysisofthebreedingand nestingoccurrencesis Db= 1.63,1.62,1.75,1.74,1.63,1.64,1.66,1.68, 1.70fortheyearsfrom2002to2010,respectively.Intheinset Dband Dfarereportedforeachyear.Valuesof Dfarereported(Additional“le 1:TableS3).

PAGE 9

Convertino etal.EcologicalProcesses 2012, 1 :9 Page9of17 http://www.ecologicalprocesses.com/content/1/1/9Therelationshipisapower-lawfunction, N ( l ) lDb, whoseexponent Dbisthefractaldimensionoftheshorebirdoccurrencepattern.Figure2areportsthepower-law relationshipforPIPLandREKNbreedingoccurrencesin 2006,andFigure2bfortheSNPLbreedingoccurrences from2002to2010.Theresultscon“rmthesupposed scale-freedistributionoftheshorebirdoccurrences.The fractaldistributionofthepredictedpatchesiscapturedby Kor cakslaw(Figure3).Thebox-countingoverestimates thefractaldimensionwithrespecttothefractaldimensionofKor cakslawasshowninConvertinoetal.(2012). Thefractaldimensionofthebox-counting( Db)is1.63, 1.85,and1.53,andthefractaldimensionofKor cakslaw ( DK)is1.47,1.70,and1.42forSNPL,PIPL,andREKN in2006,respectively(Additional“le1:TablesS2andS3). Thebox-countingenvisionsamorepessimisticscenario forthepatchsizeofshorebirds.However,asinConvertino etal.(2012)webelievethatintheabsenceofanymodeling eortbox-countingconstitutesavalidtechniquetocalculatethefractaldimensionofthemosaicofpatches.For theSNPLoccurrences,box-countingallowsustodetect the”uctuationovertimeofthefractaldimensionofthe recordednestoccurrencesandofthecoastline.Theinsets inFigure2bshowtheempiricalevidenceofthecorrelationbetween Dfand Db,andAdditional“le1:TableS3 reportsthevaluesofthefractaldimensions.Theanalysisraisesthequestionofwhetherthevariationin Dbiscausedbynatural”uctuationsofthespeciesrangeor bychangesinexternalforcingsuchasnaturaloranthropogenicstressors.Weobservethatin2004and2005the fractaldimensionshowedajumppossiblyduetothe exceptionalhurricaneseasoninthoseyears,whichaltered thepositivefeedbackbetweentropicalcyclonesandSNPL nestabundance(Convertino2011c).Thissuppositionis con“rmedbytheresultsofConvertinoetal.(2011c). Thepotentialeectofsealevelrise,oneofthemain controllingfactorsoflandcoverofcoastalhabitats,is studiedhere.Thesimulatedvariationinlandcoverclasses overtimeisperformedinSLAMM(Clough2010)for theGulfCoastofFlorida(Additional“le1:FiguresS1 andS2).Wepredictby2100adecreaseinthesalt-marsh andestuarinebeachclasses,whicharecrucialhabitatsfor PIPL,SNPL,andREKN.Wealsopredictanetdecreasein swampandinlandfreshmarshhabitats.Following”oodingpredictedtooccurafter2060,undevelopeddrylands willchangemostlyintotidal”ats,whichmayshiftinto estuarineopenwater(Additional“le1:FigureS2).We estimatea6%increaseinestuarineopenwateranda 10%increaseinoceanopenwaterfrom2006to2100. Weexpectgloballand-lossindependentofthelandcover classofabout16%withrespecttothe2msealevelrise. AvideoinAdditional“le2andFigureS2inAdditional “le1showtheevolutionoflandcoverandofthecoastlinegeomorphologyovertime.Additional“le1:Figures 100102104106108103102101100P(S>s)0.05 0.09 6 12spdf(S)a)2006 2020 2040 2020 2060 2080 2100100102104106108103102101100P(S>s)b)0.05 0.08 6 12spdf(S)100102104106108103102101100sP(S>s)6 12 0.05 0.1spdf(S)c)(m )2(x 10 m )2 6(x 10 m )2 6(x 10 m )2 6 Figure3 Kor cakslawofthepredictedsuitablepatches. The fractaldimensionofthepatchesisderivedfromthescalingexponent, = DK/ 2,oftheprobabilityofexceedanceofthepatchsize (Equation3)forSNPL,PIPL,andREKN.Theprobabilityofexceedence ofthepatchsizeisrepresentedfortheyears2006,2020,2040,2060, 2080,and2100.Theprobabilityofexceedanceiscomparedagainst thebox-countingscalinglawsfortheyears2002-2010.Additional“le 1:TableS2reportsthevaluesof DK.Theinsetsrepresentthe probabilitydensityfunctions( pdfs )ofthepatchsizethatshowa heavy-tailedbehavior.S3,S4,andS5reportthesuitabilityindexderivedfromthe predictedhabitatsuitabilitymapsusingMAXENTcorrespondingtotheyearlylandcovermaps.Thepatchesare thencalculatedusingthepatch-delineationmodelintroducedinthePatch-delineationmodelŽsectionandthe habitatsuitabilitymaps.

PAGE 10

Convertino etal.EcologicalProcesses 2012, 1 :9 Page10of17 http://www.ecologicalprocesses.com/content/1/1/9Thepower-lawstructureofthepatchsizeholdsfor everyyearsimulated(Figure3),whichprovesthescaleinvarianceofthesuitablehabitatovertime.Byusingthe maximumlikelihoodestimation(MLE)criteria,wefound thatthePareto-L evyprobabilityfunctionhasthebest“t forthepredicteddistributionofthepatchsize(Additional “le1).Kor cakslawexhibitssome“nite-sizeeectsbefore theuppertruncationandapotentiallower-cutointhe power-lawbehavior.However,thesevariationsfromthe powerlawarequitecommoninnaturalsystemsdueto the“nitenessofthevariablesampled.Thus,wecanclaim anoverallscale-invarianceofthepatchsize.Additional “le1:TableS2reportsthefractaldimensionderived fromKor cakslawfor2006,2020,2040,2060,2080,and 2100.Thescale-invarianceofthehabitatpatternsofthe SNPLwasshowninConvertinoetal.(2011b)forthepredictionofthehabitatsuitabilityin2006.Hereweshow that,giventhescale-invarianceofthepatchsize,”uctuationsinthescalingexponent = DK/ 2ofKor cakslaw occur.Webelievethatthese”uctuationsarerelatedto variationsinthelandcover,whichchangesthecoastline fractality.Thehigherthefractaldimension,thehigherthe fragmentationoftheshorebirdhabitat.Thefragmentationofthehabitatcreatessmallerpatchesforwintering andbreedingforPIPLandREKN,andforSNPL,respectively.Brownian-L evymovementsofshorebirdsmightbe thecauseforthescale-invarianceoftheoccurrencepatternsthatcanbedetectedbythebox-counting.This hasbeenprovenforothermarineanimals(Humphries etal.2010)andcolonialbirds(Jovanietal.2008).However,inthisstudywedonotreproduceanymovement ofspeciesaswebelievethatthesizeandnumberof patchesisaectedbythegeomorphologicalevolution ofthecoastline,whichinturnaectsthemovement ofshorebirds. Theworstscenarioforthevulnerabilityofshorebirds ispredictedconsideringthefractaldimension Dbfrom thebox-counting.Moreover,thebox-countingsuers fromtheriskofpotentiallyunsampledoccurrences.The Kor cakslawfractaldimension(Figure3)isbasedonthe sizeofthepredictedsuitablepatches(potentialhabitat range),whilethebox-counting(Figure2)isanapproximationthatonlycapturestherecordedoccurrences(realizedrange).Thefactthat DK = Dbforthe2006-2010 periodinwhichSNPLnestoccurrencesareavailablecon“rmsthegoodestimationofMAXENToftherealized rangeaspreviouslyfoundinConvertinoetal.(2011b). Amoreaccurateestimationofthefractaldimension thatisintermediatebetween DKand Dbisgivenbythe patchperimeter-sizescalingrelationship(Figure4).The perimeter-sizerelationshipcapturestheedgeeectsof patchesonspecies.Ingeneralshorebirdspeciespreferto liveinpatcheswhoseshapesareasregularaspossibleversushighlyirregularlyshapedpatchessuchasthepatches 0.1 1 10 100 0.010.1110p2100 PP 2006 PP 0.875 0.90 (km)b)0.1 1 10 100 0.010.1110p2100 SP 2006 SP 0.77 0.81 (km)a)0.1 1 10 100 0.010.1110pS 2100 RK 2006 RK 0.75 0.82 (km )2 (km)c) Figure4 Perimeter-sizerelationshipforSNPL,PIPL,andREKN. Perimeter-sizerelationship( p = kSD c/ 2)forthepredictedsuitable patchesoftheSNPL( a ),PIPL( b ),andREKN( c ),in2006and2100.The exponent DcfortheSNPLislistedinAdditional“le1:TableS3.determinedbyaverycomplexcoastline.Thesurvivabilityofthespeciesishigherforthoseinhabitingpatches withlargeperimetersandsimpleshapesthanforthose inhabitingpatchesofequivalentareabutcomplexshape. Thelargertheedgeeectdeterminedbythecomplexityofthepatchparameter,thelowertheprobabilityof survivalfortheindividualswithinthespecieswithinthe

PAGE 11

Convertino etal.EcologicalProcesses 2012, 1 :9 Page11of17 http://www.ecologicalprocesses.com/content/1/1/9patch.However,therearesomecasesofedgespeciesŽfor whomirregularshapesarepreferred.Inourcaseitwas observedthat DK Dc Db.Hence,theestimationofthe fractaldimensionbyusingKor cakslawforecaststhebest scenariopredictingtheleastamountoffragmentationdue tosealevelrise. Dcpredictsgreaterfragmentationthan DKbecausethefractalityofthepatchsperimeterisconsidered,butoverall DcseemsthebestestimateofthefractaldimensionbetweenKor cakslawandthebox-counting estimates. Figure5a,brespectively,showthetimeseriesofthefractaldimensionofthespecies-dependenthabitatcoastline Df(mostlybeachforSNPL,PIPL,andREKN,butalso saltmarshforthePIPL),andofthefractaldimension ofthepatches DK(fromEquation3)computedwiththe patch-delineationmodel.Additional“le1:FiguresS3,S4, andS5showthepatchesforSNPL,PIPL,andREKNin theyears2006,2020,2040,2060,2080,and2100.The majorityofpatchesarealongthebarrierislandsandparticularlyinthePanhandleregion.After2060,whensea levelsstarttorapidlyrise,aconsistentportionofthe patcheswillbefoundalongtheshoreasbarrierislands graduallydisappear.Figure5aalsoshowsthevariation inthefractaldimensionofthewholecoastlineindependentlyofthelandcoverclass.Theprobabilityof“nding apatchofsize S islowerin2100thanin2006. DKvalues aresimilarforSNPLandREKNandarehigherforPIPL (Figure5b).Thus,onaveragetherelationship DcREKN DKSNPL DKPIPLholdsforthemodeledperiod.Bigvariationsin DKPIPLareobservedparticularlyincorrespondencewithbigvariationsinthesalt-marshhabitat (Figure5a,b),whichcon“rmsthelikelihoodof“ndinga breedinggroundofPIPLinthesalt-marshhabitat(class 8containedinAdditional“le1:FigureS6)asreportedin literature(Convertinoetal.2011a)andasfoundbyour results(Additional“le1:FigureS6).In2100thefractal dimensionofREKNisverysimilartothefractaldimensionofSNPL,whilethefractaldimensionofPIPListhe highest.ThePIPLshowsthelowestprobabilityoflarge patcheswithrespecttotheothershorebirdsconsidered because DKisthehighest.Theareaunderthepowerlawdistributionofpatchesfor2100inFigure3hasa 5,3and8%negativevariationwithrespecttothearea for2006forSNPL,PIPLandREKN.Theareaunderthe curveistheoverallprobabilityof“ndingpatchesofany sizeinagivenyear.Justcomparing2006with2100is notenoughtoderiveanyconclusionaboutthespecies withthehighestpotentialriskofdecline.Thelocationand sizeofthepatchesaredeterminedbythehabitatsuitabilityat-a-pointandbyacombinationofdispersalandarea criteria(seethePatch-delineationmodelŽsection).The PipingPlover,despitehavingalargerspectrumofhabitatpreferencesthanSNPLandREKN[transitionalmarsh andsaltmarshareasarefavorableclassesasshownin 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.061.081.101.121.141.161.181.20PP SP RK 1.00 1.02 1.04 1.06 1.08 1.10 1.12 1.14 1.16 1.18 1.20 20002010202020302040205020602070208020902100 coastlinesalt marsh-waterbeach-water1.06 1.16 1.26 1.36 1.46 1.56 1.66 1.76 1.86 20002010202020302040205020602070208020902100 1.06 1.16 1.26 1.36 1.46 1.56 1.66 1.76 1.86 20002010202020302040205020602070208020902100PP SP RKa) b) c)year DfDKDKDf Figure5 Fractaldimensiontimeseriesoftheshorebirdspatches andofthecoastline. ( a )Timeseriesofthefractaldimension Dfof theentirecoastline(blueline),ofthesalt-marsh(red),andofthe beach(green)habitatcoastlines,determinedbythebox-counting algorithm.( b )Fractaldimension DKovertimeforthepatchesfor SNPL(bluedots),PIPL(red),andREKN(green)derivedfromKor caks law.Thedashedgraylines( a b )representthe95%con“dence intervaloftheestimated Dfand DK.( c )Scalingrelationshipamong thefractaldimensionofthepatchesforthethreatened,endangered, andpotentiallyat-riskshorebirdspecies(TER-s)andthefractal dimensionofthefavorablehabitatcoastlines(saltmarshforPIPL,and beachforSNPLandREKN).Theaveragespecies-independentscaling exponentis =1.67.Thegraycloud( c )representsthe95% con“denceintervalforthelinearregressionbetween DKand Df.

PAGE 12

Convertino etal.EcologicalProcesses 2012, 1 :9 Page12of17 http://www.ecologicalprocesses.com/content/1/1/9Additional“le1:FigureS6andasreportedinConvertino etal.(2011a)]seemstobeatriskduetothehighfragmentationofitshabitat.Thisisevidencedbythelarger ”uctuationof DKPIPLthan DKSNPLand DKREKN. Webelievethatitisimportanttoobservethe”uctuationsof DKovertimeforeachspecies. DKvalues ofSNPLandREKNareonaveragesteadyandincreasingovertime,respectively;thus,theprobabilityof“ndinglargepatchesfortheseshorebirdsdecreasesover timewithrespectto2006. DKofPIPLhasthelargest ”uctuations,butmostofthese”uctuationsimplyan increaseintheprobabilityof“ndinglargepatcheswith respectto2006.Nonethelesswebelievethefrequent andlargevariationinpatchesisnotagoodscenario forspecies. InFigure5cweproposeascalingrelationshipbetween thefractaldimensionofpatchesandthefractaldimensionofthehabitat-speci“ccoastline, DK Df .The relationshipholdsoveratleasttwoordersofmagnitude, fromthesmallestpatches( 0.01km2)andshortcoastlinesegmentstothelargestpatchesandthewholeFlorida Gulfcoastline.Thesamescalingexponentisobserved forSNPL,PIPL,andREKN,underliningapossiblecommonecogeomorphologicalorganizationofthelandscape undersealevelrisepressure.InFigure5c Dfischaracteristicoftheportionofcoastlineinwhichthereis asuitablehabitatforSNPL,PIPL,andREKN,whichis evidencedinAdditional“le1:FigureS6.Thecoupledevolutionofthelandcoverandhabitatpatternsmayhold cluesaboutthelinkageofgeomorphologicalandecologicalprocesses.Thescalingrelationshipbetweenthefractal dimensionsofpatchesandcoastlinecanbeapotential tooltomeasurethevulnerabilityofthespeciesinthe future.Thehighertheexponent ,thehigherthepotentialriskofdeclineofthespecies.Forsmallchangesin thecon“gurationofthecoastline,alargefragmentation ofthesuitablehabitatwouldpotentiallybeobserved.For specieswithcomparablevaluesof ,whichisthecase forSNPL,PIPL,andREKN,therangeofvaluesof DKand Dfisimportantfordetectingwhichspeciesmaybe subjectedtothemostsigni“cantchangeinthesuitable habitatpatches.Thelower DK,thehigherthelikelihood ofhavinglargepatches.Tothebestofourknowledgethis isthe“rstscalingrelationshiptobeidenti“edbetween fractaldimensionsoflandscapeandecologicalpatterns. Inthisrespectthisrelationshipbringsinsightsintothe “eldoflandscapeallometry,Žwhichisthestudyofthe possiblescalingoflandscapeandecologicalpatternsand processes.Therelationshipisbetweenfractaldimensions,whichareindicatorsthatfocusonhowmeasured quantitiesvaryasapowerofmeasurementscale,butat thesametimetherelationshiphasanallometricfocus, betweenthecoastlinecomplexityandthemagnitudeof habitatfragmentation. However,fragmentationpersedoesnotdirectlyimply lossofconnectivityamongpatches.Figure6showshow theaveragesizeofthepatches s forSNPL,PIPL,and REKNdecreaseswiththeincreaseinthefractaldimensionofthepatches.Hereweconsider DKofKor cakslaw forthefractaldimension.Atthesametimeweobserve 0 2 4 6 8 10 12 1.21.31.41.51.61.71.8sa)DKRK PP SP0 5 10 15 20 25 1.21.31.41.51.61.71.8DKcc)0 200 400 600 800 1000 1200 1.21.31.41.51.61.71.8DKNb)p(km ) 2(km) Figure6 Relationshipsamongpatchnumber,size,and connectivity,andfractaldimensionofthehabitat-speci“c coastline. s vs DK( a ), Npvs DK( b ), Npvs s ( c ),and c vs DK( d )for thethreatened,endangered,andat-riskshorebirds(TERs)considered. Thedotsarethebinaveragesover30simulationsforeachyearfor theperiod2006-2100.Thedashedlinesrepresentthe95% con“denceintervalsforthedependentvariablesconsidered.

PAGE 13

Convertino etal.EcologicalProcesses 2012, 1 :9 Page13of17 http://www.ecologicalprocesses.com/content/1/1/9anincreaseinthenumberofpatches Np.Thus,the variationinthecoastlineproducesfragmentation,rather thanshrinking,ofthesuitablehabitat.Theformerdoes notimplythelatteraserroneouslyassumedbymany theoreticalmodelsintheecologicalliterature.TheaveragesizeofthePIPLpatchesislowerthanthatforSNPL andREKN,andthehabitatforthePIPListhemostfragmented( Npisthehighestonaverage).Thisisrelated tothehighvalueof DKforthePIPLwithrespectto SNPLandREKN.Thus,althoughthevariationsin DKPIPLwouldpredictbiggerpatches,thefragmentationofthe PIPLhabitatisthegreatest.In2100thenumberofsuitablepatchesforSNPL,PIPL,andREKNispredictedtobe higherthanin2006,buttheaveragesizeofthepatches ispredictedtobesmaller(Additional“le1:TableS2).As sealevelrise(SLR)increasesthecomplexityofthecoastline,habitatpatchesmoderatelyshrinkandsplit.Onthe contrarywhenthecoastlinecomplexitydecreases,habitat patchesenlargeandcoalesce(Figure1c)asinourassumptiondepictedinFigure1b.ThePIPLseemstobethe shorebirdmostaectedbythechangesinitsbreeding habitatduetosealevelrise. Theaveragesizeandthenumberofthepatches areinverselyproportionalgiventherelationshipin Figure6a,bandasshowninAdditional“le1:FigureS7. Theaveragepatchsize s fortheshorebirdsisnotproportionaltotheaveragebodymass m aspossiblyexpected (Table1),althoughthelatterscaleswiththeaveragedispersallength.The s isforthePIPL,whileitislargerfor SNPLandREKN.Thisemphasizesthecontrollingroleof habitatgeomorphologyinshapingthepatchdistribution. ThePIPLalsodependsonthesalt-marshhabitat,whicht isoneoftheclassesmoreseriouslycompromisedbySLR. Weconsider dl,theestimatedmaximumdispersallength, inordertodeterminetheaveragenumberofconnected patches c dlconsidersrareL evy”ightsŽofindividuals ofthespeciesintheecosystem.L evy”ightsareaspecialclassofrandomwalkwithmovementdisplacements drawnfromaprobabilitydistributionwithapower-law tail(theso-calledPareto-L evydistribution),andthey giverisetostochasticprocessescloselylinkedtofractal geometryandanomalousdiusionphenomena.Because ithasthelargestmaximumdispersaldistance,theREKN hasthehighestnumberofconnectedpatches.However, forthethreeshorebirdspecies c increaseswiththe fractaldimensionofthepatches,indicatingameasureof thehabitatfragmentation.Becausewe“ndthatclimate changeisresponsibleforthesplittingofthepatches, ratherthantheirshrinking,andbecausethedispersal capabilityofspeciesisnotexpectedtochangeconsistently inthemodeledperiod,theresultseemsjusti“able.The increaseinthenumberofconnectedpatchesisexplainablebecause Npincreaseswithoutadrasticreduction inthehabitat.Theaverageconnectivityofthepredicted breedingandwinteringpatchesisanincreasingfunction ofthefractaldimensionofthepatches.TheincreasingroughnessoftheFloridacoastlineduetoclimate changeproducesalargernumberofpatcheswithsmaller dimensions.Theincreasedconnectivitywouldpotentially enhancethesurvivabilityoftheshorebirdsdespitethe decreaseintheaveragesizeofsuitablepatches.Thus,the predictedpatchpatternsfortheFloridashorebirdsare nottheworstcasescenarioinwhichboththeconnectivityandthedimensionofthepatchesarereduced.Further explanationofthelandcover,habitat,andpatchdynamics isprovidedinAdditional“le1.ConclusionsSealevelriseduetoclimatechange,beyondbeinga human-populationthreat,isshowntostronglyaect biodiversitysuchasresidentialandmigrantshorebird populationsinFlorida.Theintegratedpatch-prediction modelingframeworkproposedinthispaperconstitutesa parsimoniousbutusefulriskassessmenttoolforspecies declinewithrespecttomoreaccuratemetapopulation models.Inouropinion,theunderstandingofecogeomorphologicalprocessesatanyscaleofanalysistogetherwith thedetectionofusefulindicatorsofsuchdynamicsis oneoftheprimarygoalstoprotectbiodiversityagainst theanticipatedchangesinthelandscapeduetoclimate change.Ontheonehand,itisimpossibletoconsider,or toestimatewithlowuncertainty,allthefactorsaecting theprocessesthatgovernthedistributionofspecies(e.g., conspeci“cattractions,interspeci“ccompetition,density dependence,sexstructure,lifehistory,phenotypicplasticity,andphenologicalchangesindispersalabilityand inbreeding/winteringarearequirements),thegeomorphologicalprocesses,andthelinksandfeedbacksamong theseprocesses.Ontheotherhand,webelievethata top-downapproachofbiocomplexityisusefultodetect thefundamentaldriversoftheobservedpatternsofinterest(Schwimmer2008;NationalResearchCouncil2009; Reinhardtetal.2010).Weareawarethatmanygeomorphologicalandbiologicalprocessesarenotincorporated inthepresentedmodel;however,theuncertaintyinthe quanti“cationoftheseprocessesandtheinteractionof theseuncertaintiesmayproduceerroneousresultsinthe predictions.Theintegratedmodeliscapableofprovidingvaluablemacroscalepredictionswithrelativelyfew dataandvariables.Thus,themodelisusefulforevaluatingconservationactionsforincreasingthesurvivability ofshorebirdsinFlorida.Wearealsocon“dentthatthe proposedmodel,properlytuned,canbeappliedtomany dierentspeciesincoastalecosystemsworldwidethat arethreatenedbysealevelrise.Weanticipatefurther developmentofthismodelathigherlevelsofcomplexityandalsoforinlandsites.Thefollowingconclusionsare worthmentioning.

PAGE 14

Convertino etal.EcologicalProcesses 2012, 1 :9 Page14of17 http://www.ecologicalprocesses.com/content/1/1/9€ Ascale-freedistributionofnesting,breeding,and winteringoccurrenceswasdetectedfortheSnowy PloverinFlorida.Thescale-freedistributionwasalso foundforthewinteringoccurrencesofPipingPlover andRedKnot.Thedistributionwasderivedthrough thebox-countingtechniqueappliedtothebreeding andwinteringoccurrences,whichgivesaproxyofthe fractaldimensionofshorebirdpatches.Empirical evidenceshowsthatthefractaldimensionofthe occurrencesisstronglypositivelycorrelatedwiththe coastlinefractaldimension,whichunderlinesan ecogeomorphologicalorganization,i.e.,acouplingof ecologicalandgeomorphologicalpatterns.The powerlawheldforanyseasonoftheshorebird annualcycle,demonstratingthehighimportanceof thephysicalhabitatonspeciesprocesses. € Wepredictedbreedingandwinteringpatchesof shorebirds,simulatinglandcover(whichcomprises manycoastalwetlandtypes)andhabitatsuitabilityat theyearscalefrom2006to2100asafunctionofsea levelrise.Patcheswereidenti“edbyasetof macroecologicalcriteria,suchasarea,habitat suitability,andneighboringdistance,asafunctionof themaximumdispersal.Thedistributionofthe predictedpatchsizewasKor cakslaw,whose exponentishalfofthefractaldimensionofthe patches.Wevalidatedthemodelbypredictingthe observedpatch-sizedistributionandpatchpatterns from2002to2010wheredatawereavailable.Wealso investigatedtheperimeter-sizerelationshipfor estimatingthefractaldimensionofthepatchesata higherlevelofcomplexitybecauseofthecalculation oftheperimeter.Thefractaldimensionprovidedby theperimeter-sizerelationshipprovidedamedian estimatebetweenthevaluesderivedfromKor caks lawandthebox-countingdistribution.Kor cakslaw providedthemostoptimisticscenarioof fragmentationinwhichtheprobabilityof“nding largepatcheswasthehighest,whilethebox-counting providedthemostpessimisticscenario.Hence,the perimeter-arearelationshipissuggestedasthebest methodtocalculatethefractaldimensionofthe mosaicofhabitatpatches. € TherobustnessofthePareto-L evydistributionofthe patchsizewasveri“edforpredictionsofpatches from2006to2100.Thus,thescale-invarianceofthe patchpatternsholdsintimedespitethestrong in”uenceofsealevelrise.Thismayberelatedtoa sortofsimulatedbiologicalresilienceŽofspeciesto theexternalchanges(Folkeetal.2004)byassuming invarianthabitatareaanddispersalrequirement. Scale-freehabitatpatternshaveproventobethe mostresilienttoexternalstressorsinpreviousstudies (Ke“etal.2011).Thus,theshapeofthepatch-size probabilityandthefractaldimensionwhenthis probabilityisapowerlawcanbeusefulindicatorsto estimatethedegreeofstressŽofcoastalecosystems. Furtherresearchisanticipatedtounderstandwhen andhowthepatch-sizeprobabilitydeviatesfroma Pareto-L evybehavior.Thefragmentation,whichis proportionaltothefractaldimensionofthe habitat-speci“ccoastline,variedconsiderablyover timeandinparticularforthePipingPlover.However, theriskofextirpationin2100forSNPL,PIPL,and REKNwasnothighwithrespectto2006.Wenote thatthecomparisonbetween“nalandinitialyears riskshouldnotbetheonlycomparisoninevaluating theriskofdeclineofaspecies.Theoveralltrendof thefractaldimensioninthemodeledperiodhastobe evaluatedaswell. € Ascalingrelationshipwasfoundbetweenthefractal dimensionsofthepatchesandofthehabitat-speci“c coastline.Thescalingexponentofthisrelationship appearstobespecies-independentfortheshorebirds considered.Furtherresearchisneededtoexplorethe conditionsofuniversality(species-andecosystemwise)ofthisrelationship,whichmayberelatedtothe speciesconsidered.The”uctuationinthefractal dimensionofthecoastlinecanbeassumedtobea valuableecologicalindicatorforassessingvariationin patchpatternsofbreedingandwinteringshorebirds. € Wedemonstratedthathabitatloss,fragmentation, andconnectivityarethreeseparateconcepts. Althoughthesevariablesarecloselylinkedtoeach other,theircausalityisnottrivial.Fortheshorebirds studied,thepredictedfragmentationwascoupled withhabitatlosswhiletheconnectivityincreased.The factthatthepatches,evenifsmaller,wereconnected isanextremelypositivefactorthatensuresdispersal andgene”ow;thus,theconnectivityofpatches enhancesthesurvivabilityofshorebirds.Birth,death, anddispersalprocessesofaspeciescanovercomethe habitat-losseectandadecreaseintheaveragesize ofpatches.Yet,alowermetapopulationriskof extirpationexistsifinterpatchmigrationisallowed (KindvallandPetersson2000).However,adecrease intheaveragepatchsizecanpotentiallyincrease intra-speciescompetitionforforaging(Ritchie1998) anddecreasecarryingcapacity.Apossibleoptimal ecogeomorphologicalstateofthecoastalecosystem maybecharacterizedbythesmallestfractal dimensionofthecoastlinethatmaximizesthe compactnessofthesuitablepatches.This con“gurationalsominimizesthefractaldimensionof thepatches.Thehighestentropyofthiscon“guration maytranslateintothesmallestenergyexpenditureof thespeciesthatinhabitthehabitat,forexample,for foragingandbreedingactivities.Theentropyof

PAGE 15

Convertino etal.EcologicalProcesses 2012, 1 :9 Page15of17 http://www.ecologicalprocesses.com/content/1/1/9geomorphologicallandforms(Nievesetal.2010)may, infact,behighlycorrelatedwiththescale-invariance ofecologicalpatternssuchasspecies-patchpatterns.EndnotessubmittedtoEcologicalProcessesSpecialIssueWetlandsInaComplexWorldŽ,GuestEditor:Dr.Matteo ConvertinoAdditional“les Additional“le1:AdditionalMethods,AdditionalResultsand Discussion,AdditionalTables,andAdditionalFigure. Additional“le2:VideoS1. PredictedlandcoverbySLAMMfrom2006to 2100. Abbreviations SNPL:SnowyPlover;PIPL:PipingPlover;REKN:RedKnot;TER:threatened, endangered,andatrisk;SLAMM:SeaLevelAectingMarshesModel;SLR:sea levelrise; Df:fractaldimensionofthecoastline(frombox-counting); Db:fractal dimensionofthebreedingandwinteringoccurrences(frombox-counting); DK:fractaldimensionofthepatches(fromKor cakslaw); Dc:fractaldimension ofthepatches(fromperimeter-sizerelationship); S :patch-size; p :patch perimeter; P ( hs ) :habitatsuitabilityscore;SI:suitabilityindex; Sp:minimum populationpatch-size; Sb / w:minimumbreeding/winteringpatch-size; hr : home-range; hrd :home-rangedistance; dl:maximumdispersallength;MLE: maximumlikelihoodestimation. Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. Authorscontributions MCdesignedthestudy,managedandanalyzedthedata,wrotethemodel (box-countingandpatchdelineationmodel),developedthetheory,and wrotethemanuscript.ABassistedinmakingthecalculationsandanalysis,and helpedinwritingthemanuscript.GAKandRMCparticipatedinthehabitat suitabilitymodelingframeworkandreviewedthemanuscript.ILsupervised thewholework,andreviewedthemanuscriptbyprovidingapracticalangle tothisresearchforeectiveenvironmentalmanagement.Allauthorsreadand approvedthe“nalmanuscript. Authorsinformation MCisResearchScientistattheUniversityofFlorida,Gainesville,anda ContractoroftheEngineeringResearchandDevelopmentCenteroftheUS ArmyCorpsofEngineersattheRiskandDecisionScienceTeam.ABiscurrently a“nancialanalystatFrontierAirlines.ABgothisB.ScandM.Sc.fromMIT,Civil andEnvironmentalEngineeringprogram.ABperformedhisresearch internshipattheRiskandDecisionScienceTeaminthesummerof2011.GAK andRMCareAssociateandProfessorattheUniversityofFlorida,Gainesville, respectively.IListeamleaderoftheRiskandDecisionScienceTeamofthe EngineeringResearchandDevelopmentCenteroftheUSArmyCorpsof Engineers. Acknowledgements ThisresearchwassupportedbytheUSDepartmentofDefense,throughthe StrategicEnvironmentalResearchandDevelopmentProgram(SERDP),Project SI-1699.M.C.acknowledgesthefundingofprojectDecisionandRiskAnalysis ApplicationsEnvironmentalAssessmentandSupplyChainRisksŽforhis researchattheRiskandDecisionScienceTeam.Thecomputationalresources oftheUniversityofFloridaHigh-PerformanceComputingCenter(http://hpc. u”.edu)arekindlyacknowledged.TheauthorscordiallythankDr.RAFisher (EngineeringResearchandDevelopmentCenteroftheUSArmyCorpsof Engineers)andtheEglinAirForceBasepersonnelfortheirhelpinobtaining thedataandfortheusefulinformationaboutthebreedinginformationof SNPL.TyndallAirForceBaseandFloridaWildlifeCommissionarealso gratefullyacknowledgedfortheassistancewiththedata.WethankM.L. Chu-Agor(currentlyattheCenterofEnvironmentalSciences,Departmentof BiologyandEarthandAtmosphericSciences,SaintLouisUniversity,St.Louis, MO)forhercomputationaleortwithSLAMMattheUniversityofFlorida. PermissionwasgrantedbytheUSACEChiefofEngineerstopublishthis material.Theviewsandopinionsexpressedinthispaperarethoseofthe individualauthorsandnotthoseoftheUSArmyorothersponsor organizations. Authordetails1DepartmentofAgriculturalandBiologicalEngineering-IFAS,Universityof Florida,Gainesville,FL,USA.2ContractorattheRiskandDecisionScience Team,EnvironmentalLaboratory,EngineerResearchandDevelopment Center,USArmyCorpsofEngineers,Concord,MA,USA.3FloridaClimate Institute,UF-FSU,c/oFrazierRogersHall,Gainesville,FLUSA.4Departmentof CivilandEnvironmentalEngineering,MassachusettsInstituteofTechnology, Cambridge,MA,USA.5DepartmentofEarth,Atmospheric,andPlanetary Science,MassachusettsInstituteofTechnology,Cambridge,MA,USA.6Risk andDecisionScienceTeam,EnvironmentalLaboratory,EngineerResearchand DevelopmentCenter,USArmyCorpsofEngineers,Concord,MA,USA.7DepartmentofEngineeringandPublicPolicy,CarnegieMellonUniversity, Pittsburgh,PA,USA. Received:11June2012Accepted:11September2012 Published:30October2012 References Aiello-LammensME,Chu-AgorML,ConvertinoM,FischerRA,LinkovI,Resit AkcakayaH(2011)TheimpactofsealevelriseonsnowyploversinFlorida: integratinggeomorphological,habitat,andmetapopulationmodels. GlobalChangeBiol17:3644…3654.doi:10.1111/j.1365-2486.2011.02497.x AiroldiL(2003)Eectsofpatchshapeinintertidalalgalmosaics:rolesofarea, perimeteranddistancefromedge.MarBiol143:639…650 AllianceFS(2010)FloridaPanhandleShorebirdWorkingGroup.Tech.rep FWC/AudubonFlorida/USFWS.http://www.”shorebirdalliance.org/ resources-pages/maps.html Audubon(2006)Americastoptenmostendangeredbirds.Tech.rep.National AudubonSociety,Newyork BakP(1999)Hownatureworks:thescienceofself-organizedcriticality. Springer-Verlag,Berlin BaldassarriA,SapovalB,FelixS(2012)Anumericalretro-actionmodelrelates rockycoasterosiontopercolationtheory.Arxiv.http://arxiv.org/pdf/1202. 4286v1.pdf BanavarJR,ColaioriF,FlamminiA,MaritanA,RinaldoA(2001)Scaling, optimalityandlandscapeevolution.JStatPhys104:1…33 BattyM,Longley,P(eds)(1994)Fractalcities.AcademicPress,SanDiego BonabeauE,DagornL,FreonP(1999)Scalinginanimalgroup-size distributions.ProcNatAcadSci96:4472…4477.doi:10.1073/pnas.96.8.4472 BuldyrevS,DokholyanN,ErramilliS,HongM,KimJ,MalescioG,StanleyH (2003)Hierarchyinsocialorganization.PhysStatMechAppl330(3-4): 653…659.doi:10.1016/j.physa.2003.09.041 BurneyC(2009)Floridabeach-nestingbirdreport,2005-2008.Tech.rep. FloridaFishandWildlifeConservationCommission,Tallahassee.http:// www.”shorebirdalliance.org/pdf/2005-2008 FWC BNB Report.pdf C-CAP(2009)Coastalchangeanalysisprogramregionallandcover.Tech.rep. NOAA,WashingtonDC.http://www.csc.noaa.gov/digitalcoast/data/ ccapregional/ Chu-AgorM,Mu noz-CarpenaR,KikerG,EmanuelssonA,LinkovI(2011) Exploringsealevelrisevulnerabilityofcoastalhabitatsusingglobal sensitivityanduncertaintyanalysis.EnvironModelSoftware26(5):593…604 CloughJ(2006)ApplicationofSLAMM4.1toninesitesinFlorida.Warren PinnacleConsulting,Waits“eld,VT.http://warrenpinnacle.com/prof/ SLAMM/NWF SLAMM FLORIDA 2-16-2006.doc CloughJS(2010)TheSeaLevelAectingMarshesModel.Tech.rep.Warren PinnacleConsulting,Waits“eld,VT.http://warrenpinnacle.com/prof/ SLAMM6/SLAMM6 Technical Documentation.pdf ConvertinoM,KikerG,Chu-AgorM,Mu noz-CarpenaR,MartinezC, Aiello-LammensM,Akc akayaH,FisherR,LinkovI(2010)Integrated

PAGE 16

Convertino etal.EcologicalProcesses 2012, 1 :9 Page16of17 http://www.ecologicalprocesses.com/content/1/1/9modelingtomitigateclimatechangeriskduetosea-levelriseofimperiled shorebirdsonFloridacoastalmilitaryinstallations.In:LinkovI,BridgesT (eds)Climatechange:globalchangeandlocaladaptation.Springer, Dordrecht ConvertinoM,DonoghueJ,Chu-AgorM,KikerG,Munoz-CarpenaR,FischerR, LinkovI(2011a)Anthropogenicrenourishmentfeedbackonshorebirds:a multispeciesBayesianperspective.EcolEng37:1184…1194. doi:10.1016/j.ecoleng.2011.02.019 ConvertinoM,KikerG,Munoz-CarpenaR,Chu-AgorM,FischerR,LinkovI (2011b)Scale-andresolution-invarianceofsuitablegeographicrangefor shorebirdmetapopulations.EcolComplexity8(4):364…376. doi:10.1016/j.ecocom.2011.07.007 ConvertinoM,ElsnerJ,Mu noz-CarpenaR,KikerG,FisherR,LinkovI(2011c)Do tropicalcyclonesshapeshorebirdhabitatpatterns?Biogeoclimatologyof SnowyPloversinFlorida.PLoSONE6(1):p1 ConvertinoM,SiminiF,CataniF,KikerG(2012)Fromriverbasinstoelephants tobacteriacolonies:aggregate-sizespectrumofanimateandinanimate species.PLoSONE(inpress) delBarrioG,HarrisonP,BerryP,ButtN,SanjuanM,PearsonR,DawsonT(2006) Integratingmultiplemodellingapproachestopredictthepotential impactsofclimatechangeonspeciesdistributionsincontrastingregions: comparisonandimplicationsforpolicy.EnvironSciPolicy9(2):129…147. doi:10.1016/j.envsci.2005.11.005 EhrlichPR,LevinSA(2005)Theevolutionofnorms.PLoSBiol3(6):e194. doi:10.1371/journal.pbio.0030194 ElliottSmithE,HaigS(2004)Pipingplover(charadriusmelodus).In:PooleA (ed)TheBirdsofNorthAmericaOnline.CornellLabofOrnithology,Ithaca, doi:10.2173/bna.2 F-DEP,StateGeologicalMapFDEP-GEO(2001).Tech.rep.,FloridaDepartment ofEnvironmentalProtection(datafrom,FGDL,UniversityofFlorida GEOPLANCenter).http://www.dep.state.”.us/geology/gisdatamaps/ state geo map.htm FallonF(2005)Petitiontolisttheredknot(caladriscanutusrufa)asendangered andrequestforemergencylistingundertheendangeredspeciesact.Tech. rep.DelawareRiverkeeperNetwork,AmericanLittoralSociety,Delmarva OrnithologicalSociety,DelawareChapteroftheSierraClub,NewJersey AudubonSociety.http://www.fws.gov/northeast/redknot/riverkeeper.pdf FolkeC,CarpenterS,WalkerB,ScheerM,ElmqvistT,GundersonL,HollingC (2004)Regimeshifts,resilienceandbiodiversityinecosystem management.AnnuRevEcolEvolSyst35:557…581 FWC(2010)Floridabeach-nestingbirdswebsite.Tech.rep.FloridaFishand WildlifeConservationCommission.http://legacy.myfwc.com/bnb/ HarringtonB(2001)Redknot(calidriscanutus).PooleA(ed)TheBirdsofNorth AmericaOnline.CornellLabofOrnithology,Ithaca,doi:10.2173/bna.563 HelzerC,JelinskiD(1999)EcolApp9:1448…1458 HimesJ,DouglassN,PrunerR,CroftA,SeckingerE(2006)Statusand DistributionofSnowyPloverinFlorida,2006.Tech.rep.FloridaWildlife ConservationCommission.http://www.”shorebirdalliance.org/pdf/ Himes Douglass-2006 SNPL Report.pdf HumphriesN,etal.(2010)EnvironmentalcontextexplainsL evyandBrownian movementpatternsofmarinepredators.Nature465(7301):1066…1069. doi:10.1038/nature09116 ImreAR,BogaertJ(2004)Thefractaldimensionasameasureofthequalityof habitats.ActaBiotheoretica52:41…56 JovaniR,TellaJ(2007)Fractalbirdnestdistributionproducesscale-freecolony sizes.ProcRSocB274:2465…2469.doi:10.1098/rspb.2007.0527 JovaniR,SerranoD,Urs uaE,TellaJL,AdlerFR(2008)Truncatedpowerlaws revealalinkbetweenlow-levelbehavioralprocessesandgrouping patternsinacolonialbird.PLoSONE3:1992…+. doi:10.1371/journal.pone.0001992 K e“S,RietkerkM,AladosCL,PueyoY,PapanastasisVP,ElaichA,deRuiterPC (2007)Spatialvegetationpatternsandimminentdeserti“cationin Mediterraneanaridecosystems.Nature449:213…217. doi:10.1038/nature06111 Ke“S,RietkerkM,RoyM,FrancA,deRuiterP,PascualM(2011)Robustscaling inecosystemsandthemeltdownofpatchsizedistributionsbefore extinction.EcolLett14(1):29…35.doi:10.1111/j.1461-0248.2010.01553.x KindvallO,PeterssonA(2000)onsequencesofmodellinginterpatchmigration asafunctionofpatchgeometrywhenpredictingmetapopulation extinctionrisk.EcolModell129(1):101…109 KlemasV,DobsonJ,FergusonR,HaddadK(1993)Acoastallandcover classi“cationsystemforthenoaacoastwatchchangeanalysisproject.J CoastalRes9(3):862…872 KorcakJ(1940)Deuxtypesfondamentauxdedistributionstastique.BullInst IntStar30:295…307 LamonteK,DouglassN(2002)StatusandDistributionofSnowyPloverin Florida,2002.Tech.rep.FloridaWildlifeConservationCommission. http://www.”shorebirdalliance.org/pdf/Lamonte Douglass2002 SNPL Report.pdf LeyrerJ,SpaansB,CamaraM,PiersmaT(2006)Smallhomerangesandhigh site“delityinredknots(calidrisc.canutus)winteringonthebancdArguin, mauritania.JOrnithology147(2):376…384.doi:10.1007/s10336-005-0030-8 MajkaD,JennessJ,BeierP(2007)Corridordesigner:Arcgistoolsfordesigning andevaluatingcorridors.CorridorDesigner.http://corridordesign.org MandelbrotB(1967)Howlongisthecoastofbritain?Statisticalself-similarity andfractionaldimension.Science156(3775). doi:10.1126/science.156.3775.636 Mandelbrot,B(1982)TheFractalGeometryofNature.W.H.Freeman,NewYork MasettiR,FagherazziS,MontanariA(2008)Applicationofabarrierisland translationmodeltothemillennial-scaleevolutionofSandKey,Florida. ContinentalShelfRes28:1116…1126.doi:10.1016/j.csr.2008.02.021 MoraisPA,OliveiraEA,Ara ujoNAM,HerrmannHJ,AndradeJS(2011)Fractality oferodedcoastlinesofcorrelatedlandscapes.PhysRevE84:016,102. doi:10.1103/PhysRevE.84.016102 MullerS,Mu noz-CarpenaR,KikerG(2010)Modelrelevance:frameworksfor, exploringthecomplexity-sensitivity-uncertaintytrilemma.LinkovI,Bridges T(eds)ClimateChange:GlobalChangeandLocalAdaptation.NATObook, Amsterdam NationalResearchCouncil,NAS(ed)(2009)ANewBiologyforthe21st Century.TheNationalAcademiesPress,Washington,DC NievesV,WangJ,BrasR,WoodE(2010)Maximumentropydistributionsof scale-invariantprocesses.PhysRevLett105(11):118,701+. doi:10.1103/PhysRevLett.105.118701 NikoraV,PearsonC,ShankarU(1999)Scalingpropertiesinlandscapepatterns: NewZealandExperience.LandscapeEcol14:17…33 PageG,StenzelL,WarrinerJ,PatonP(2009)Snowyplover(charadrius alexandrinus).In:CornellILabofOrnithology,PooleA(eds)TheBirdsof NorthAmericaOnline,vol.5.doi:10.2173/bna.154,http://bna.birds.cornell. edu/bna/species/154 PaolaC,LeederM(2011)Environmentaldynamics:Simplicityversus complexity.Nature469:38…39.doi:10.1038/469038a PascualM,ManojitR,GuichardF,FlierlG(2002)Cluster-sizedistributions: signaturesofself-organizationinspatialecologies.PhilTransRSocLondB 357:657…666.doi:10.1098/rstb.2001.0983 PascualM,RoyM,LaneriK(2011)Simplemodelsforcomplexsystems: exploitingtherelationshipbetweenlocalandglobaldensities.TheorEcol 4(2):211…222 PatonP,EdwardsT(1996)Factorsaectinginterannualmovementsofsnowy plovers.TheAuk113(3):534…543.doi:10.2307/4088973 PhillipsS,AndersonR,SchapireR(2006)Maximumentropymodelingof speciesgeographicdistributions.EcolModell190(3-4):231…259. doi:10.1016/j.ecolmodel.2005.03.026 PhillipsSJ,MiroslavD(2008)Modelingofspeciesdistributionswithmaxent: newextensionsandacomprehensiveevaluation.Ecography31(2): 161…175 PrunerR(2010)Assessinghabitatselection,reproductiveperformance,and theaectsofanthropogenicdisturbanceoftheSnowyPloveralongthe FloridaGulfcoast.Mastersthesis,UniversityofFlorida,Gainesville,USA ReinhardtL,JerolmackD,CardinaleBJ,VanackerV,WrightJ(2010)Dynamic interactionsoflifeanditslandscape:feedbacksattheinterfaceof geomorphologyandecology.EarthSurfProcessesLandforms35:78…101. doi:10.1002/esp.1912 RinaldoA,DietrichWE,RigonR,VogelGK,Rodrlguez-LturbeI(1995) Geomorphologicalsignaturesofvaryingclimate.Nature374:632…635. doi:10.1038/374632a0 RitchieM(1998)Scale-dependentforagingandpatchchoiceinfractal environments.EvolutionaryEcol12:309…330 ScanlonTM,CaylorKK,LevinSA,Rodriguez-IturbeI(2007)Positivefeedbacks promotepower-lawclusteringofKalaharivegetation.Nature449: 209…212.doi:10.1038/nature06060

PAGE 17

Convertino etal.EcologicalProcesses 2012, 1 :9 Page17of17 http://www.ecologicalprocesses.com/content/1/1/9SchneiderD,TellaJ,Scalingtheory:applicationtomarineornithology(2002). Ecosystems5:736…748.doi:10.1007/s10021-002-0156-y SchwimmerRA(2008)Atemporalgeometricanalysisoferodingmarsh shorelines:canfractaldimensionsberelatedtoprocesses.JCoastalRes 1:152…158 SeaveyJR,GilmerB,McGarigalKM(2010)Eectofsealevelriseonpiping plover(charadriusmelodus)breedinghabitat.BiolConserv. doi:10.1016/j.biocon.2010.09.017 Sol eR,Bascompte,J(Eds)(2006)Self-organizationinComplexEcosystems. PrincetonUniversityPress,Princeton,NJ StenzelL,WarrinerJ,WarrinerJ,WilsonK,BidstrupF,PageG(1994) Long-distancebreedingdispersalofSnowyPloversinwesternNorth America.JAnimalEcol63:887…902 StenzelL,PageG,WarrinerJ,WarrinerJ,GeorgeD,EysterC,RamerB,Neuman K(2007)Survivalandnataldispersalofjuvenilesnowyplovers( Charadrius alexandrinus )incentralcoastalCalifornia.TheAuk124(3):1023…1036 USFWS(2009)PipingPlover(Charadriusmelodus)5-YearReview:S ummaryand Evaluation.Tech.rep.,USFishandWildlifeService.http://www.fws.gov/northeast/ endangered/PDF/Piping Plover “ve year review and summary.pdf USGS(2010)NationalElevationDataset.Tech.rep.,UnitedStateGeological Survey.http://ned.usgs.gov/ USGS-FWS(2009)Datafromthe2006,InternationalPipingPloverCensus (IPIPLC),Tech.rep.,UnitedStateGeologicalSurveyandUSFWS.http:// www.”shorebirdalliance.org/pdf/Elliot-Smith Haig-2006 PIPL Report.pdf WarrinerJS,WarrinerJC,PageGW,StenzelLE(1986)Matingsystemand reproductivesuccessofasmallpopulationofpolygamousSnowyPlovers. WilsonBull98:15…37 doi:10.1186/2192-1709-1-9 Citethisarticleas: Convertino etal. : Shorebirdpatchesas“ngerprintsof fractalcoastline”uctuationsduetoclimatechange. EcologicalProcesses 2012 1 :9. Submit your manuscript to a journal and bene“ t from:7 Convenient online submission 7 Rigorous peer review 7 Immediate publication on acceptance 7 Open access: articles freely available online 7 High visibility within the “ eld 7 Retaining the copyright to your article Submit your next manuscript at 7 springeropen.com


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E9Z29AEZH_9HJTM6 INGEST_TIME 2013-03-05T20:25:24Z PACKAGE AA00013662_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


!DOCTYPE art SYSTEM 'http:www.biomedcentral.comxmlarticle.dtd'
ui 2192-1709-1-9
ji 2192-1709
fm
dochead Research
bibl
title
p Shorebird patches as fingerprints of fractal coastline fluctuations due to climate change
aug
au id A1 ca yes snm Convertinofnm Matteoinsr iid I1 I2 I3 email mconvertino@ufl.edu
A2 BockelieAdamI4 I5 bockelie@mit.edu
A3 Kikermi AGregorygkiker@ufl.edu
A4 Muñoz-CarpenaRafaelcarpena@ufl.edu
A5 LinkovIgorI6 I7 Igor.Linkov@usace.army.mil
insg
ins Department of Agricultural and Biological Engineering-IFAS, University of Florida, Gainesville, FL, USA
Contractor at the Risk and Decision Science Team, Environmental Laboratory, Engineer Research and Development Center, US Army Corps of Engineers, Concord, MA, USA
Florida Climate Institute, UF-FSU, c/o Frazier Rogers Hall, Gainesville, FL, USA
Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA
Department of Earth, Atmospheric, and Planetary Science, Massachusetts Institute of Technology, Cambridge, MA, USA
Risk and Decision Science Team, Environmental Laboratory, Engineer Research and Development Center, US Army Corps of Engineers, Concord, MA, USA
Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, PA, USA
source Ecological Processes
issn 2192-1709
pubdate 2012
volume 1
issue 1
fpage 9
url http://www.ecologicalprocesses.com/content/1/1/9
xrefbib pubid idtype doi 10.1186/2192-1709-1-9
history rec date day 11month 6year 2012acc 1192012pub 30102012
cpyrt 2012collab Convertino et al.; licensee Springer.note This is an Open Access article distributed under the terms of the Creative Commons Attribution License(http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
kwdg
kwd Land cover change
Coastal wetlands
Coastline complexity
Fractal dimension
Habitat suitability
Patches
Sea level rise
abs
sec
st
Abstract
Introduction
The Florida coast is one of the most species-rich ecosystems in the world. This paper focuses on the sensitivity of the habitat of threatened and endangered shorebirds to sea level rise induced by climate change, and on the relationship of the habitat with the coastline evolution. We consider the resident Snowy Plover (it Charadrius alexandrinus nivosus), and the migrant Piping Plover (Charadrius melodus) and Red Knot (Calidris canutus) along the Gulf Coast of Mexico in Florida.
Methods
We analyze and model the coupled dynamics of habitat patches of these imperiled shorebirds and of the shoreline geomorphology dictated by land cover change with consideration of the coastal wetlands. The land cover is modeled from 2006 to 2100 as a function of the A1B sea level rise scenario rescaled to 2 m. Using a maximum-entropy habitat suitability model and a set of macroecological criteria we delineate breeding and wintering patches for each year simulated.
Results
Evidence of coupled ecogeomorphological dynamics was found by considering the fractal dimension of shorebird occurrence patterns and of the coastline. A scaling relationship between the fractal dimensions of the species patches and of the coastline was detected. The predicted power law of the patch size emerged from scale-free habitat patterns and was validated against 9 years of observations. We predict an overall 16% loss of the coastal landforms from inundation. Despite the changes in the coastline that cause habitat loss, fragmentation, and variations of patch connectivity, shorebirds self-organize by preserving a power-law distribution of the patch size in time. Yet, the probability of finding large patches is predicted to be smaller in 2100 than in 2006. The Piping Plover showed the highest fluctuation in the patch fractal dimension; thus, it is the species at greatest risk of decline.
Conclusions
We propose a parsimonious modeling framework to capture macroscale ecogeomorphological patterns of coastal ecosystems. Our results suggest the potential use of the fractal dimension of a coastline as a fingerprint of climatic change effects on shoreline-dependent species. Thus, the fractal dimension is a potential metric to aid decision-makers in conservation interventions of species subjected to sea level rise or other anthropic stressors that affect their coastline habitat.
meta classifications classification WCW subtype theme_series_title type BMC Wetlands in a complex worldtheme_series_editor Matteo Convertinobdy
Introduction
Florida coastline-dependent species are characterized by one of the highest extirpation risks in the world because of sea level rise and increase in tropical cyclone activity (Convertino et al. abbr bid B16 2010;
B19 2011c) due to climate change. The Snowy Plover (Charadrius alexandrinus nivosus; SNPL hereafter) is a residential shorebird of Florida listed as threatened at the state level. The Piping Plover (Charadrius melodus; PIPL hereafter) is federally designated as threatened, and it migrates mostly from the North Atlantic coasts of the USA and Canada to Florida where it winters for 3 months on average (Elliott Smith and Haig B23 2004). The Red Knot (Calidris canutus; REKN hereafter) is designated as threatened in New Jersey and is federally listed as a potential “at risk” species. REKN uses the Florida Gulf beaches as stop-over areas for about 3 weeks during its migration between South America and North America’s Big Lakes region and Atlantic coast (Harrington B28 2001). This is considered as the wintering period of the REKN in Florida. An understanding of the spatial distribution of the suitable habitat patches for these shorebirds, their controlling factors, and how these factors are affected by sea level rise is fundamentally important for adopting efficient conservation strategies. An understanding of linkages between the coupled evolution of landforms and ecological patterns is a crucial topic due to the evidence that these patterns are tightly linked. Biocomplexity approaches (Mandelbrot B44 1982; Rinaldo et al. B60 1995; Banavar et al. B7 2001; Pascual et al. B53 2002; Schneider and Tella B63 2002; Buldyrev et al. B10 2003; del Barrio et al. B21 2006; Solé and Bascompte B66 2006; Scanlon et al. B62 2007), despite being accused of adopting simplified biological models (Paola and Leeder B52 2011), are capable of reproducing macroscale patterns of complex phenomena and of developing indicators, such as the probability of the patch size (Mandelbrot 1982; Bonabeau et al. B9 1999; Jovani and Tella B33 2007; Kéfi et al. B35 2007; Jovani et al. B34 2008; Convertino et al. B20 2012), that are useful for assessing ecosystem health (Kefi et al. B36 2011). One of ecology’s main goals is to detect from observed patterns, such as species occurrence patterns, the organizational rules of species in stationary and evolving ecosystems. Many theories have been proposed to explain the formation of clustered patterns of species in nature. Conspecific attraction, environmental heterogeneities, and food availability have been claimed—alone or together—to be the motivation for the formation of habitat patches in which individuals of a species coexist in colonies. An optimal search theory, the so-called Lévy-flight foraging hypothesis (or predator-prey-food resource dynamics), predicts that predators should adopt search strategies known as Lévy flights where prey is sparse and distributed unpredictably. However, Humphries et al. (B31 2010) showed that Brownian movement is sufficiently efficient for locating abundant prey. This theory explains the clustered patterns of resources in landscapes that may be different from the pattern of species occurrence. Neither the Lévy-flight foraging hypothesis nor Brownian movement model address the linkages of biota with landscape forms and their evolution, which is, in our opinion, one of the main missing points.The colony size of seabirds (Schneider and Tella 2002), colonial birds (Jovani and Tella 2007), and many other other animals (Bonabeau et al. 1999) has been found to follow a power-law distribution. Analogous scale-free distributions have been detected for bacteria colonies (Buldyrev et al. 2003), for species in complex ecosystems (Solé and Bascompte 2006; Convertino et al. 2012), and also for man-made systems such as cities (Batty and Longley B8 1994). The ubiquity of the power-law structure for the probability of the patch size in aggregation phenomena of natural and human systems suggests the existence of universal self-organization principles (Pascual et al. 2002; Solé and Bascompte 2006). The scaling exponent of the power-law distribution of the aggregate size was proven to be the fractal dimension of the pattern analyzed (Mandelbrot 1982; Convertino et al. 2012). The word “aggregate” is a general word for indicating the assemblage of individuals with similar or identical features in a landscape. In the presence of a power law for the probability distribution of the aggregate size, the occurrence patterns are scale-free, indicating that the patterns are invariant at different scales of observations (Convertino et al. 2012). The concept of fractal dimension was introduced by Mandelbrot analyzing the coastline of Britain at different scales (Mandelbrot B43 1967). The work disseminated the use of fractal analysis first in geomorphology (Morais et al. B46 2011; Baldassarri et al. B6 2012) and later to a variety of sciences from biology to engineering (Bak B5 1999). Nonetheless, all these theories, models, and empirical findings have rarely considered any potential effect of slow or abrupt change in the exogenous factors on the heterogenous habitat in which species live. Only recently it was proven quantitatively that ecosystems exhibit variations in the probability distribution of the patch size due to anthropically and naturally driven changes in the environmental variables (Kefi et al. 2011). For example, desertification of water-controlled ecosystems produces a decrease in the fractal dimension of vegetation patches, or in extreme cases, a shift from the power law to exponential distribution of the patch size (Kéfi et al. 2007; Scanlon et al. 2007; Kefi et al. 2011). Climate change scenarios tested in temperate/continental regions depicted an overall decrease in the fractal dimension of patches in time for many different taxa (Barrio et al. 2006). For colonial birds the variation in the fractal dimension of the patches was clearly related to the fluctuations in the population abundance due to interspecies competition (Jovani et al. 2008). In geomorphology the variation in the fractal dimension was used as the signature of the persisting climate over landscapes. For example, the association between landscape evolution and climate has been assessed for river basin ecosystems in Rinaldo et al. (1995). However, none of the previous studies linked the fractal dimension of two ecosystems’ patterns in time (e.g., of geomorphological and ecological patterns) resulting from linked processes. Here we verify for the first time, to the best of our knowledge, that the fractality of the coastline is clearly linked to the habitat patches of shoreline-dependent birds in their breeding and wintering seasons.We hypothesize that sea level rise may increase the complexity of the coastline and that such complexity determines fragmentation of the habitat of species. We assume scale invariance of the patches, which is also detectable by the analysis of the shorebird occurrences. We consider a breeding shorebird (Snowy Plover) and wintering shorebirds (Piping Plover and Red Knot) in Florida to quantify the potential effect of sea level rise on resident and migrant species. For the Snowy Plover the nesting season is usually considered part of the breeding season; thus, our model’s input considers the SNPL breeding and nesting occurrences simultaneously. Furthermore, observations indicate that nesting, breeding, and wintering areas for SNPL fall within the same range (Convertino et al. B17 2011a). Wintering occurrences of SNPL are thus considered together with breeding occurrences.An integrated ecogeomorphological modeling approach is adopted to predict the viability from 2006 to 2100 of threatened, endangered, and at risk (TER) shorebirds (SNPL, PIPL, and REKN) along the Gulf Coast of Florida as a function of the increasing sea level rise due to climate change. We rescale to 2 m the Intergovernmental Panel on Climate Change (IPCC A1B) scenario described in Chu-Agor et al. (B13 2011) and model the ecosystem at a 120 m spatial resolution. We predict land cover change with the Sea Level Affecting Marshes Model smcaps SLAMM (Clough B15 2010)] which is a geomorphological model at low-medium level of complexity. SLAMM considers coastal wetland types such as swamp, cypress swamp, mangrove, and salt marsh (Additional file supplr sid S1 1: Figure S1). The habitat model predicts the habitat suitability for breeding and wintering through a maximum entropy principle approach (MaxEnt) (Phillips and Miroslav B57 2008) as a function of the recorded species occurrences in the breeding and wintering season, the predicted land cover, and a geology layer. MaxEnt is an ecological model at low level of complexity. The land cover and habitat simulations are produced in Aiello-Lammens et al. (B1 2011). Finally, in this paper a patch-delineation model is introduced to predict the yearly habitat patches for a set of biological constraints imposed on the habitat suitability maps. We assume the stationarity of the habitat patterns at the year scale and absence of biological adaptation of species to climate change. The fractal dimension of the patches is derived by three independent methods: (i) box-counting for the observed occurrences; (ii) probability distribution of the patch size [“Korčak’s law” (Korcak B39 1940; Mandelbrot 1982)]; and (iii) perimeter-area relationship for the predicted patches. We assume that these three methods produce very close estimates of the fractal dimension of the whole mosaic of patches as shown in Convertino et al. (2012).
suppl
Additional file 1
text
b Additional Methods, Additional Results and Discussion, Additional Tables, and Additional Figure.
file name 2192-1709-1-9-S1.pdf
Click here for file
The power-law distribution of the patch size is verified by almost a decade (2002–2010) of historical observations of the species. Thus, the patch-delineation model is validated against these observations from 2002 to 2010. The coupled ecogeomorphological organization is shown by the correspondence in time of the fractal dimensions of the habitat-specific coastline and of the predicted patches. The fractal dimension of the habitat-specific coastline, along with habitat loss and population abundance, is demonstrated to greatly influence the number and size of the patches, which are related to habitat loss and population abundance. Although the fragmentation of the habitat (which is proportional to the fractal dimension of the patches) is predicted to fluctuate considerably in this century, the risk of extirpation of the species analyzed is not drastically increased because the connectivity of the patches is predicted to increase. The Piping Plover is the species with the largest fluctuation in the number and size of patches. We believe the research presented in this paper constitutes a contribution to the emerging field of biogeosciences, which explores the interface between biology and the geosciences and attempts to understand the interrelated functions of landscapes and biological systems across multiple spatial and temporal scales. We are aware of the existence of many other complex ecogeomorphological processes that are not included in our modeling effort. However, parsimonious models such as the model presented here can capture large-scale patterns while bypassing small-scale details (Ehrlich and Levin B22 2005; Pascual et al. B54 2011). These models can be tested against other more biologically realistic models to fully explore the linkages among various environmental changes, geomorphological dynamics, and biodiversity patterns. We anticipate that further research will explore this issue of process complexity versus model complexity, model relevance, and model uncertainty, which can be synthesized as a “modeling trilemma” (Muller et al. B47 2010).This paper is organized as follows. The “Methods” section describes the shorebird data and the study site and explains the models used in this study and the theoretical characterization of patches. The “Results and discussion” section reports the main results with a broad discussion of figures and how these results are interpreted considering our assumptions. The “Conclusions” section reports the most important conclusions, implications for management, and further research efforts. Additional files 1 and S2 2 are provided to support our main result.
Additional file 2
Video S1. Predicted land cover by SLAMM from 2006 to 2100.
2192-1709-1-9-S2.avi
Click here for file
Methods
Site description and biogeographical variables
The white fine-sand beaches of the Florida coast of the Gulf of Mexico constitute the habitat of the whole Florida SNPL population. The SNPL population in Florida is distributed along about 80% of the Florida Panhandle and along about 20% of the Florida Peninsula (Lamonte and Douglass B40 2002; Himes et al. B30 2006; Burney B11 2009; Pruner B58 2010) (Figure figr fid F1 1a). The Florida Peninsula and the Atlantic coasts are the main wintering grounds for the migratory PIPL and REKN, which seem less constrained than the SNPL by the mineralogical properties of the beach substrate captured by the geology layer (Convertino et al. 2010;
B18 2011b). The land cover, which includes many wetland types from C-CAP (B12 2009) is represented in Figure S1 of the Additional file 1, and the geology (F-DEP B24 2001) characterizes the mineralogical substrate of each land cover class (Additional file 1: Figure S6) (Convertino et al. 2011b). In 2006 the PIPL Panhandle-Peninsula and Atlantic populations were 38 and 33%, respectively, of the total migrant PIPL population in Florida. The REKN Panhandle-Peninsula and Atlantic populations were 55 and 20%, respectively, of the total migrant population in Florida. The International Piping Plover Census in 2006 supported the field sampling of SNPL, PIPL, and REKN (USGS-FWS B71 2009; FWC B27 2010; Alliance B3 2010). The 2006 wintering occurrences in Florida are the data used in this study for PIPL and REKN. For the SNPL, data of breeding and nesting occurrences are also available from 2002 to 2010 and are provided year by year by the Florida Wildlife Commission. These occurrences are used to verify the assumption of scale-invariance of SNPL occurrence patterns over time with the box-counting. However, despite the availability of SNPL data from 2002 to 2010, we construct the habitat suitability model with the 2006 SNPL occurrences alone in order to be consistent with the 2006 NOAA land cover (C-CAP 2009) and the 2006 PIPL and REKN occurrences. The geology and the elevation from USGS (USGS B70 2010; Convertino et al. 2010;
2011b) are used in the habitat suitability model and in the land cover model, respectively (Aiello-Lammens et al. 2011; Convertino et al. 2010;
2011b).We consider PIPL and REKN in the same geographic domain where the full range of the SNPL occurs in order to perform a simultaneous interspecies assessment of the habitat use and extirpation risk of the three species (Figure 1a). Thus, only the Panhandle-Peninsula region was considered in this study. The SNPL is our main interest because its year-round presence in the Florida coastal ecosystem makes this species potentially more vulnerable than PIPL and REKN. Dispersal among the Panhandle and Peninsula SNPL populations has been observed but not quantified. Population subdivision of the SNPL has not been observed; thus, we can adopt the same habitat and dispersal criteria for the whole population. Population subdivision, for example, can be caused by geographic barriers or disturbances [e.g., renourishment (Convertino et al. 2011a)] that interfere with the dispersal. The reduction in dispersal is reported to reduce gene flow and increase genetic drift of independent subpopulations in the long-term. However, this is not the case for the SNPL population in Florida despite the weak interchange of individuals between Panhandle and Peninsula (Aiello-Lammens et al. 2011).Habitat area and dispersal data for SNPL are mostly from Aiello-Lammens et al. (2011) but also from Page et al. (B51 2009), Patons and Edwards (B55 1996), Stenze et al. (B67 1994; B68 2007), Warriner et al. (B72 1986). Aiello-Lammens et al. 2011synthesized the biological data and the metapopulation modeling effort of this research for the SNPL. Information is gathered also from field ecologists working on this project [i.e., Dr. R.A. Fischer (Engineering Research and Development Center, US Army Corps of Engineers) and Mrs. A. Pruner (Florida Park Service)]. For PIPL, habitat and dispersal data are from Audubon (B4 2006), Seavey et al. B65 2010, and USFWS (B69 2009), and for REKN, data are from Fallon (B25 2005) and Leyrer et al. (B41 2006). For a more detailed description of the site under study we refer the reader to Convertino et al. (2011b).
Box-counting algorithm
The characterization of the occurrence patterns of breeding and wintering occurrences and of the coastline is performed using the “box-counting” method. For the SNPL the occurrence pattern of nesting occurrences was observed to be a self-similar pattern (Convertino et al. 2012); thus, the box-counting method is suitable to predict how this pattern changes with the scale of analysis. The box-counting analysis consists of calculating, for grids of different box-side lengths, the number of boxes that contain the object under study. Adjacent boxes constitute an approximation of the real patches at each resolution. The algorithm can be applied to both point and line patterns. The box-counting is performed over eight orders of magnitude in a logarithmic scale of the box-side length, from l
sub
o(1)=565km (which corresponds to the box “B1” in Figure 1), which is approximatively the width of Florida, to l
o(5000)=11.3×10sup −3 km (Figure 1). We indicate with l
o(i) the length of the box side at resolution “o=i” from i=1,…,5000 where the increment from one resolution to another is 11.3×10−2km, which is slightly smaller than the average home range of the SNPL (Table tblr tid T1 1). The order of magnitude is relative to the scales of analysis (extent) investigated by the box-counting, while the resolution is related to the grids chosen for the box-counting. The number N(l)of boxes of size l needed to cover the pattern of occurrences (which is generally a fractal set) follows a power law,
display-formula M1
m:math 2192-1709-1-9-i1 xmlns:m http:www.w3.org1998MathMathML m:mi N
m:mo (
l
)
=
m:msub
m:mrow
N
m:mn 0
m:mspace width 2.77695pt
m:msup
l

D
,
where D≤d, and d is the dimension of the space (usually d=1,2,3). D is also known as the Minkowski-Bouligand dimension, Kolmogorov capacity, or Kolmogorov dimension, or simply the box-counting dimension and is an estimate of the Hausdorff dimension (Mandelbrot 1982). The fractal dimension for 1−d objects is associated with the Hurst exponent H such that D=2−H(Mandelbrot 1982; Bak 1999). The values of the Hurst exponent vary between 0 and 1, with higher values indicating a smoother trend, less volatility, and less roughness of the analyzed pattern (Mandelbrot 1982). We indicate the fractal dimension of the breeding and wintering occurrences with D
b
and the fractal dimension of the coastline with D
f
derived from box-counting analysis. Both fractal dimensions are determined by the box-counting method. The fractal dimension of the coastline is calculated also for each land cover class that is a species-specific habitat for the species considered (Figure 1b). Many land cover classes are coastal wetland types (Additional file 1: Figure S1).
fig Figure 1caption Box-counting algorithm
Box-counting algorithm. (a) Representation of the the box-counting algorithm applied to the 2006 occurrences of the Snowy Plover (SNPL), Piping Plover (PIPL), and Red Knot (REKN), for eight orders of magnitude (in a logarithmic scale), which corresponds to 5000 resolutions of the box-counting grid. In this example at the resolution of box B5 the number of boxes in which there is at least one occurrence is N(B5)=6. (b) Box-counting example applied to the whole coastline, to the habitat-specific coastline (e.g., beach, salt-marsh), and to other land cover classes as in Convertino et al. (2011b). Many coastal wetland types are included in the land cover, such as swamp, cypress swamp, mangrove, and salt marsh. The shaded grid cells in (a) and (b) have at least one species occurrence or a coastline segment at the represented resolution. Two coastline configurations are presented: the first for high values of Dfand DK(c), the second for low values of Dfand DK(d). The patches presented in green are connected because their neighboring distance is lower than the maximum dispersal length dl.
graphic 2192-1709-1-9-1
table
Table 1
Macroecological parameters of the patch-delineation model, and biological data estimated from the literature
tgroup align left cols 4
colspec colname c1 colnum 1 colwidth 1*
center c2 2
c3 3
c4
thead valign top
row
entry
SNPL
PIPL
REKN
rowsep
Model parameters
tfoot
S
p
and S
b/w
are the minimum population and breeding/wintering area, respectively. d
l
is the estimated maximum dispersal length. 〈nd〉is the neighborhood distance in the breeding season for Snowy Plover (SNPL), and in the wintering season for Piping Plover (PIPL) and Red Knot (REKN). hr and hrd are the home range and the home range distance estimated considering the breeding regions for SNPL, PIPL, and REKN. m is the average body mass. Data are taken from Burney (2009), Himes et al. (2006), Lamonte and Douglass (2002), Pruner (2010), Aiello-Lammens et al. (2011). Model parameters are assumed considering data and calibrating the patch-delineation model on the observed patch size in 2006 derived from the box-counting.
tbody
S
p
(km2)
0.03
0.04
0.06
S
b/w
(km2)
0.01
0.02
0.04
d
l
(km)
8
12
20
Data
hr (km2)
0.016
2.20
10÷15
hrd (km)
0.12
1.48÷2.40
3.10
〈nd〉 (km)
0.72
0.96
1.44
m (g)
38÷40
50÷60
180÷200
Land-cover model
The land cover is predicted year by year by using the Sea Level Affecting Marshes Model (SLAMM) (Clough B14 2006; Chu-Agor et al. 2011) starting from the year 2006 to 2100. These simulations are performed in Aiello-Lammens et al. (2011) and Convertino et al. (2010) to which we refer the reader for more details. The domain of the model is extended inland for about 10 km from the coastline (Convertino et al. 2011a;
2011b) (black region along the coast in Figure 1, box B1). We consider the predicted inundation distance in 2100 (∼ 9 km) for a range of [1, 2] m sea level rise (SLR) adding 1 km to consider the uncertainty in the estimation of the flooding distance. The initial condition is the 2006 land cover from NOAA (Klemas et al. B38 1993). The NOAA land cover classes are changed into SLAMM land cover classes for modeling purposes. SLAMM requires us to group the classes of land cover into model classes. The conversion is reported in Convertino et al. (2011b). The SLAMM model also requires the elevation and slope as input variables. The modeled domain is divided into seven regions (Additional file 1: Figure S1) with distinct historical tidal and SLR trends. Each region is characterized by a unique set of values for the 26 input parameters (Additional file 1: Table S1) related to tide, accretion, sedimentation, and erosion processes. The value of the parameters is derived from the available literature and previous efforts of this research (Chu-Agor et al. 2011). In this effort of modeling the land cover, we do not consider any geomorphological feedback between landforms and climate change that is expected to occur with global warming. All our assumptions are the same as those in Chu-Agor et al. (2011) and Convertino et al. (2010). Also we do not consider any possible barrier island shifting because that is reported to occur over a time period much longer than our predictions (Masetti et al. B45 2008).
Habitat suitability model
The employed habitat suitability model is MaxEnt (Phillips et al. B56 2006; Phillips and Miroslav 2008), which is one of the most diffused models in species distribution modeling. MaxEnt is a model based on the principle of maximum entropy that predicts continuous habitat suitability maps of potential species occurrence under a set of selected environmental variables. The environmental variables that are necessary and sufficient for calculating the habitat suitability are the land cover translated into SLAMM classes (Chu-Agor et al. 2011; Convertino et al. 2011a;
2011b) and the USGS geology layer (Convertino et al. 2011a;
2011b) at a resolution of 120 m. The resolution 120 m is the home-range distance of the SNPL (Table 1). Such distance is sufficient to capture not only the spatial variability of habitat preferences of SNPL, but also that of PIPL and REKN, whose home-range distance are much larger than that of SNPL. The habitat suitability at-a-point (i.e., for each pixel of the modeled domain) can be considered as a proxy to find SNPL, PIPL, and REKN in the breeding and wintering season. The prior probabilities of occurrence are calculated in MaxEnt using the recorded shorebird occurrences constrained to the environmental variables. The occurrences are nest and breeding occurrences for SNPL, and the adult occurrences for PIPL and REKN. Thus, for PIPL and REKN the habitat suitability refers to the suitability for wintering as in Convertino et al. (2011a). No absences are required in MaxEnt. Then the posterior probabilities of occurrence are based on the prior probabilities given the change in the land cover modeled year by year by the land cover model. A regularization parameter that controls the fit of the predicted suitability to the real occurrence data is assumed to be equal to one. Non-randomly placed pseudoabsences are used to improve the predictions, and 25% of the occurrences are taken as a training sample (Convertino et al. 2011a;
2011b). The predicted habitat suitability maps represent the average of over 30 replicates for each year to reduce the uncertainty of the predictions. The habitat suitability is calculated with 10,000 random background points. Background points are a subset of points of the domain over which the Bayesian inference between the recorded species occurrences, pseudoabsences, and environmental layers is determined.We assign a biological interpretation to the predicted habitat suitability score, P(hs), which is the probability at-a-point of finding a breeding and/or a wintering ground. Breeding and wintering grounds are suitable sites for the SNPL as a function of the season considered, and wintering grounds are suitable sites for the PIPL and REKN. We define the suitability index (SI) as a metric from 0 to 100 that captures the quality of the breeding and/or wintering habitat for the species. The higher the SI the larger the biological spectrum of functions performed by the species in that habitat. Hence, P(hs) is also a surrogate of habitat use during the breeding and wintering seasons of the species considered. In fact, it is legitimate to assume that habitat use increases with habitat quality. Every pixel of the HS maps is classified into five SI categories: SI=100 [for 0.8≤P(hs)≤1] is considered the best habitat with the highest survival and/or reproductive success; SI=80 [for 0.6≤P(hs)<0.8] is typically associated with successful breeding and/or wintering; SI=60 [for 0.2≤P(hs)<0.6] is associated with consistent use for breeding and wintering; SI=30 [for 0.2 Patch-delineation model
Below we define a criterion for delineating breeding and wintering patches for SNPL, PIPL and REKN, respectively. A patch is defined when the following criteria simultaneously hold:
M2
2192-1709-1-9-i2 m:mtext for
m:mfenced separators open { close
m:mtable columnalign
m:mtr
m:mtd
SI

60
[ ]
i.e., for
P
(
hs
)

0
.
2

P(X) within a neighborhood distance
mathvariant italic nd

d
l
Minimum population patch size

S
p
Minimum breeding/wintering patch size

S
b
/
w
.
The species-dependent values for the three parameters required for the patch identification are reported in Table 1. The values of biological data in Table 1 are used only to support the choice of model parameters. The model parameters are calibrated to reproduce a patch-size distribution as close as possible to the box-counting distribution of occurrences in 2006. The model with this set of parameters was validated against the patch-size distributions from 2002 to 2010 estimated by the box-counting. We define breeding patch as an area large enough to at least occasionally support a single breeding pair through courtship and rearing of young to dispersal age (Majka et al. B42 2007). A population patch is defined as an area large enough to support breeding for 10 years or more, even if the patch is isolated from interaction with other populations of the species (Majka et al. 2007). Since population-wide data are lacking for these breeding and population area requirements, we assumed that a population patch is at least two times larger than a breeding patch. For the SNPL these patches contain certain nesting patches. The minimum population and breeding/wintering patch areas are estimated from the literature available and by expert knowledge of the field biologists involved in the sampling campaigns performed for this study (see Burney 2009; Himes 2006; Lamonte and Douglass 2002; Pruner 2010). S
p
and S
b/w
are the minimum population and breeding/wintering area, respectively, and are proportional to the estimated home range. The minimum breeding/wintering area is the minimum area that will support breeding and wintering activity of the shorebirds. The home range hr and the home-range distance hrd (the square root of the hr) are values estimated considering the breeding regions for SNPL, PIPL, and REKN. We assume that S
p
and S
b/w
for PIPL, and REKN are much smaller than hr because they refer to the wintering period of these shorebirds in Florida. For REKN, S
p
is also reduced due to the habitat limitation and the close coexistence with SNPL in the same habitat. Patches are considered connected if their neighboring distance is equal to or smaller than d
l
, which is the maximum dispersal length. Figure 1c,d shows an example of patches that are connected because their reciprocal distance is lower than d
l
. These plots also represent our assumption that coastline complexity affects patch distribution. The average neighborhood distance 〈nd〉 is the average dispersal of the species. 〈nd〉 is higher than hrd for the SNPL due to the higher local dispersal ability estimated from recent surveys (Himes et al. 2006; Pruner 2010). For PIPL and REKN, 〈nd〉is smaller than hrd because the reported hrd refers to their breeding range in northern states in the USA and Canada. In the winter season PIPL and REKN migrate to Florida, and their dispersal distance is observed to be smaller. Within the neighborhood distance a subpopulation can be assumed to be panmictic. A panmictic population is one in which all individuals are potential partners. It is usually estimated from the foraging distance of an animal species. In a more abstract way the neighborhood distance is the glue of all the suitable patches. In a particle physics analogy, it describes the Brownian motion of individuals within a larger species group. Thus, by using d
l
, which is the maximum dispersal, as a criterion in the model, foraging is certain to be considered within patches. Our model considers an upper estimate of the patch size for all the shorebirds considered. m is the average body mass, which is used to discuss some results. We assume the same biological parameters for the SNPL Panhandle and Peninsula as in Aiello-Lammens et al. (2011).
Probability distribution of the patch size
The probability of exceedance of the patch size is known in literature as Korčak’s law (Korcak 1940; Nikora et al. B50 1999), which is expressed by:
M3
2192-1709-1-9-i3 P
(
S

s
)
=
c
S

ε
F
( )
m:mfrac
s
s
c
,
where c is a constant, F is a homogeneity function that depends on a characteristic size s
c
, and ε=D
K
/2 is the scaling exponent (Korcak 1940; Mandelbrot 1982). D
K
is the fractal dimension of the patches. The probability of exceedance exhibits a power-law behavior. The probability distribution of the patch size for the predicted patches was used to validate the patch-delineation model against the box-counting estimates on the real occurrences from 2002 to 2010. The fit of the predicted distribution of patches is performed using a Maximum Likelihood Estimation technique (MLE), which is described in the Additional file 1.
Perimeter-area relationship
The scaling relationship between the perimeter p and the size S of the patches:
M4
2192-1709-1-9-i4 p
=
k
S
D
c
/
2
,
determines the fractal dimension of the mosaic of patches, which considers the fractality of the patch edge. Here we indicate the fractal dimension D
c
, which is derived from the same predicted patches of the introduced patch model (see the “Patch-delineation model” section) but also considers their perimeters. Because Korčak’s law (Korcak 1940) considers only the size of the patches, the perimeter-area scaling law has been considered as a more precise tool for measuring the fractal dimension. In literature the ratio p/Sis adopted to measure the quality of the patches for population survivability, that is, the likelihood of surviving in a suitable patch (Helzer and Jelinski B29 1999; Airoldi B2 2003; Imre and Bogaert B32 2004). In general the higher the ratio p/Sthe less suitable the patch area for the species, and the higher the ratio p/S, the higher the fractal dimension D
c
.
Results and discussion
The relationship between the number of cells occupied by shorebird occurrences, N(l), and the length of the side of the box, l, at each scale of analysis is shown in Figure F2 2. The relationship is a power-law function, N(l)∼l
D
b
, whose exponent D
b
is the fractal dimension of the shorebird occurrence pattern. Figure 2a reports the power-law relationship for PIPL and REKN breeding occurrences in 2006, and Figure 2b for the SNPL breeding occurrences from 2002 to 2010. The results confirm the supposed scale-free distribution of the shorebird occurrences. The fractal distribution of the predicted patches is captured by Korčak’s law (Figure F3 3). The box-counting overestimates the fractal dimension with respect to the fractal dimension of Korčak’s law as shown in Convertino et al. (2012). The fractal dimension of the box-counting (D
b
) is 1.63, 1.85, and 1.53, and the fractal dimension of Korčak’s law (D
K
) is 1.47, 1.70, and 1.42 for SNPL, PIPL, and REKN in 2006, respectively (Additional file 1: Tables S2 and S3). The box-counting envisions a more pessimistic scenario for the patch size of shorebirds. However, as in Convertino et al. (2012) we believe that in the absence of any modeling effort box-counting constitutes a valid technique to calculate the fractal dimension of the mosaic of patches. For the SNPL occurrences, box-counting allows us to detect the fluctuation over time of the fractal dimension of the recorded nest occurrences and of the coastline. The insets in Figure 2b show the empirical evidence of the correlation between D
f
and D
b
, and Additional file 1: Table S3 reports the values of the fractal dimensions. The analysis raises the question of whether the variation in D
b
is caused by natural fluctuations of the species range or by changes in external forcing such as natural or anthropogenic stressors. We observe that in 2004 and 2005 the fractal dimension showed a jump possibly due to the exceptional hurricane season in those years, which altered the positive feedback between tropical cyclones and SNPL nest abundance (Convertino 2011c). This supposition is confirmed by the results of Convertino et al. (2011c).
Figure 2Box-counting scaling-law in time
Box-counting scaling-law in time. (a) Power-law N(l)=N0l−Dbderived from the box-counting algorithm applied to the occurrences of PIPL (black dots) and REKN (green) in 2006 and to the whole Florida Gulf coastline. In the inset the schematized Florida coastline is evaluated at different box sizes. (b) Box-counting algorithm applied to the 2002-2010 occurrence of the SNPL. The fractal dimension derived from the analysis of the breeding and nesting occurrences is Db=1.63, 1.62, 1.75, 1.74, 1.63, 1.64, 1.66, 1.68, 1.70 for the years from 2002 to 2010, respectively. In the inset Dband Dfare reported for each year. Values of Dfare reported (Additional file 1: Table S3).
2192-1709-1-9-2
Figure 3Korčak’s law of the predicted suitable patches
Korčak’s law of the predicted suitable patches. The fractal dimension of the patches is derived from the scaling exponent, ε=DK/2, of the probability of exceedance of the patch size (Equation 3) for SNPL, PIPL, and REKN. The probability of exceedence of the patch size is represented for the years 2006, 2020, 2040, 2060, 2080, and 2100. The probability of exceedance is compared against the box-counting scaling laws for the years 2002-2010. Additional file 1: Table S2 reports the values of DK. The insets represent the probability density functions (pdfs) of the patch size that show a heavy-tailed behavior.
2192-1709-1-9-3 The potential effect of sea level rise, one of the main controlling factors of land cover of coastal habitats, is studied here. The simulated variation in land cover classes over time is performed in SLAMM (Clough 2010) for the Gulf Coast of Florida (Additional file 1: Figures S1 and S2). We predict by 2100 a decrease in the salt-marsh and estuarine beach classes, which are crucial habitats for PIPL, SNPL, and REKN. We also predict a net decrease in swamp and inland fresh marsh habitats. Following flooding predicted to occur after 2060, undeveloped drylands will change mostly into tidal flats, which may shift into estuarine open water (Additional file 1: Figure S2). We estimate a 6% increase in estuarine open water and a 10% increase in ocean open water from 2006 to 2100. We expect global land-loss independent of the land cover class of about 16% with respect to the 2 m sea level rise. A video in Additional file 2 and Figure S2 in Additional file 1 show the evolution of land cover and of the coastline geomorphology over time. Additional file 1: Figures S3, S4, and S5 report the suitability index derived from the predicted habitat suitability maps using MaxEnt corresponding to the yearly land cover maps. The patches are then calculated using the patch-delineation model introduced in the “Patch-delineation model” section and the habitat suitability maps.The power-law structure of the patch size holds for every year simulated (Figure 3), which proves the scale-invariance of the suitable habitat over time. By using the maximum likelihood estimation (MLE) criteria, we found that the Pareto-Lévy probability function has the best fit for the predicted distribution of the patch size (Additional file 1). Korčak’s law exhibits some finite-size effects before the upper truncation and a potential lower-cutoff in the power-law behavior. However, these variations from the power law are quite common in natural systems due to the finiteness of the variable sampled. Thus, we can claim an overall scale-invariance of the patch size. Additional file 1: Table S2 reports the fractal dimension derived from Korčak’s law for 2006, 2020, 2040, 2060, 2080, and 2100. The scale-invariance of the habitat patterns of the SNPL was shown in Convertino et al. (2011b) for the prediction of the habitat suitability in 2006. Here we show that, given the scale-invariance of the patch size, fluctuations in the scaling exponent ε=D
K
/2 of Korčak’s law occur. We believe that these fluctuations are related to variations in the land cover, which changes the coastline fractality. The higher the fractal dimension, the higher the fragmentation of the shorebird habitat. The fragmentation of the habitat creates smaller patches for wintering and breeding for PIPL and REKN, and for SNPL, respectively. Brownian-Lévy movements of shorebirds might be the cause for the scale-invariance of the occurrence patterns that can be detected by the box-counting. This has been proven for other marine animals (Humphries et al. 2010) and colonial birds (Jovani et al. 2008). However, in this study we do not reproduce any movement of species as we believe that the size and number of patches is affected by the geomorphological evolution of the coastline, which in turn affects the movement of shorebirds.The worst scenario for the vulnerability of shorebirds is predicted considering the fractal dimension D
b
from the box-counting. Moreover, the box-counting suffers from the risk of potentially unsampled occurrences. The Korčak’s law fractal dimension (Figure 3) is based on the size of the predicted suitable patches (potential habitat range), while the box-counting (Figure 2) is an approximation that only captures the recorded occurrences (realized range). The fact that D
K
≅D
b
for the 2006-2010 period in which SNPL nest occurrences are available confirms the good estimation of MaxEnt of the realized range as previously found in Convertino et al. (2011b). A more accurate estimation of the fractal dimension that is intermediate between D
K
and D
b
is given by the patch perimeter-size scaling relationship (Figure F4 4). The perimeter-size relationship captures the edge effects of patches on species. In general shorebird species prefer to live in patches whose shapes are as regular as possible versus highly irregularly shaped patches such as the patches determined by a very complex coastline. The survivability of the species is higher for those inhabiting patches with large perimeters and simple shapes than for those inhabiting patches of equivalent area but complex shape. The larger the edge effect determined by the complexity of the patch parameter, the lower the probability of survival for the individuals within the species within the patch. However, there are some cases of “edge species” for whom irregular shapes are preferred. In our case it was observed that D
K
≤D
c
≤D
b
. Hence, the estimation of the fractal dimension by using Korčak’s law forecasts the best scenario predicting the least amount of fragmentation due to sea level rise. D
c
predicts greater fragmentation than D
K
because the fractality of the patch’s perimeter is considered, but overall D
c
seems the best estimate of the fractal dimension between Korčak’s law and the box-counting estimates.
Figure 4Perimeter-size relationship for SNPL, PIPL, and REKN
Perimeter-size relationship for SNPL, PIPL, and REKN. Perimeter-size relationship (inline-formula 2192-1709-1-9-i5 p
=
k
S
m:msubsup
D
c

/
2
) for the predicted suitable patches of the SNPL (a), PIPL (b), and REKN (c), in 2006 and 2100. The exponent Dcfor the SNPL is listed in Additional file 1: Table S3.
2192-1709-1-9-4 Figure F5 5a,b respectively, show the time series of the fractal dimension of the species-dependent habitat coastline D
f
(mostly beach for SNPL, PIPL, and REKN, but also salt marsh for the PIPL), and of the fractal dimension of the patches D
K
(from Equation 3) computed with the patch-delineation model. Additional file 1: Figures S3, S4, and S5 show the patches for SNPL, PIPL, and REKN in the years 2006, 2020, 2040, 2060, 2080, and 2100. The majority of patches are along the barrier islands and particularly in the Panhandle region. After 2060, when sea levels start to rapidly rise, a consistent portion of the patches will be found along the shore as barrier islands gradually disappear. Figure 5a also shows the variation in the fractal dimension of the whole coastline independently of the land cover class. The probability of finding a patch of size S is lower in 2100 than in 2006. D
K
values are similar for SNPL and REKN and are higher for PIPL (Figure 5b). Thus, on average the relationship
2192-1709-1-9-i6
D
c
REKN

D
K
SNPL

D
K
PIPL
holds for the modeled period. Big variations in
2192-1709-1-9-i7
D
K
PIPL
are observed particularly in correspondence with big variations in the salt-marsh habitat (Figure 5a,b), which confirms the likelihood of finding a breeding ground of PIPL in the salt-marsh habitat (class 8 contained in Additional file 1: Figure S6) as reported in literature (Convertino et al. 2011a) and as found by our results (Additional file 1: Figure S6). In 2100 the fractal dimension of REKN is very similar to the fractal dimension of SNPL, while the fractal dimension of PIPL is the highest. The PIPL shows the lowest probability of large patches with respect to the other shorebirds considered because D
K
is the highest. The area under the power-law distribution of patches for 2100 in Figure 3 has a 5, 3 and 8% negative variation with respect to the area for 2006 for SNPL, PIPL and REKN. The area under the curve is the overall probability of finding patches of any size in a given year. Just comparing 2006 with 2100 is not enough to derive any conclusion about the species with the highest potential risk of decline. The location and size of the patches are determined by the habitat suitability at-a-point and by a combination of dispersal and area criteria (see the “Patch-delineation model” section). The Piping Plover, despite having a larger spectrum of habitat preferences than SNPL and REKN [transitional marsh and salt marsh areas are favorable classes as shown in Additional file 1: Figure S6 and as reported in Convertino et al. (2011a)] seems to be at risk due to the high fragmentation of its habitat. This is evidenced by the larger fluctuation of
2192-1709-1-9-i8
D
K
PIPL
than
2192-1709-1-9-i9
D
K
SNPL
and
2192-1709-1-9-i10
D
K
REKN
.
Figure 5Fractal dimension time series of the shorebirds patches and of the coastline
Fractal dimension time series of the shorebirds patches and of the coastline. (a) Time series of the fractal dimension Dfof the entire coastline (blue line), of the salt-marsh (red), and of the beach (green) habitat coastlines, determined by the box-counting algorithm. (b) Fractal dimension DKover time for the patches for SNPL (blue dots), PIPL (red), and REKN (green) derived from Korčak’s law. The dashed gray lines (a, b) represent the 95% confidence interval of the estimated Dfand DK. (c) Scaling relationship among the fractal dimension of the patches for the threatened, endangered, and potentially at-risk shorebird species (TER-s) and the fractal dimension of the favorable habitat coastlines (salt marsh for PIPL, and beach for SNPL and REKN). The average species-independent scaling exponent is γ= 1.67. The gray cloud (c) represents the 95% confidence interval for the linear regression between DKand Df.
2192-1709-1-9-5 We believe that it is important to observe the fluctuations of D
K
over time for each species. D
K
values of SNPL and REKN are on average steady and increasing over time, respectively; thus, the probability of finding large patches for these shorebirds decreases over time with respect to 2006. D
K
of PIPL has the largest fluctuations, but most of these fluctuations imply an increase in the probability of finding large patches with respect to 2006. Nonetheless we believe the frequent and large variation in patches is not a good scenario for species.In Figure 5c we propose a scaling relationship between the fractal dimension of patches and the fractal dimension of the habitat-specific coastline, D
K
∼D
f
γ
. The relationship holds over at least two orders of magnitude, from the smallest patches (∼ 0.01km2) and short coastline segments to the largest patches and the whole Florida Gulf coastline. The same scaling exponent is observed for SNPL, PIPL, and REKN, underlining a possible common ecogeomorphological organization of the landscape under sea level rise pressure. In Figure 5c
D
f
is characteristic of the portion of coastline in which there is a suitable habitat for SNPL, PIPL, and REKN, which is evidenced in Additional file 1: Figure S6. The coupled evolution of the land cover and habitat patterns may hold clues about the linkage of geomorphological and ecological processes. The scaling relationship between the fractal dimensions of patches and coastline can be a potential tool to measure the vulnerability of the species in the future. The higher the exponent γ, the higher the potential risk of decline of the species. For small changes in the configuration of the coastline, a large fragmentation of the suitable habitat would potentially be observed. For species with comparable values of γ, which is the case for SNPL, PIPL, and REKN, the range of values of D
K
and D
f
is important for detecting which species may be subjected to the most significant change in the suitable habitat patches. The lower D
K
, the higher the likelihood of having large patches. To the best of our knowledge this is the first scaling relationship to be identified between fractal dimensions of landscape and ecological patterns. In this respect this relationship brings insights into the field of “landscape allometry,” which is the study of the possible scaling of landscape and ecological patterns and processes. The relationship is between fractal dimensions, which are indicators that focus on how measured quantities vary as a power of measurement scale, but at the same time the relationship has an allometric focus, between the coastline complexity and the magnitude of habitat fragmentation.However, fragmentation per se does not directly imply loss of connectivity among patches. Figure F6 6 shows how the average size of the patches 〈s〉for SNPL, PIPL, and REKN decreases with the increase in the fractal dimension of the patches. Here we consider D
K
of Korčak’s law for the fractal dimension. At the same time we observe an increase in the number of patches N
p
. Thus, the variation in the coastline produces fragmentation, rather than shrinking, of the suitable habitat. The former does not imply the latter as erroneously assumed by many theoretical models in the ecological literature. The average size of the PIPL patches is lower than that for SNPL and REKN, and the habitat for the PIPL is the most fragmented (N
p
is the highest on average). This is related to the high value of D
K
for the PIPL with respect to SNPL and REKN. Thus, although the variations in
2192-1709-1-9-i11
D
K
PIPL
would predict bigger patches, the fragmentation of the PIPL habitat is the greatest. In 2100 the number of suitable patches for SNPL, PIPL, and REKN is predicted to be higher than in 2006, but the average size of the patches is predicted to be smaller (Additional file 1: Table S2). As sea level rise (SLR) increases the complexity of the coastline, habitat patches moderately shrink and split. On the contrary when the coastline complexity decreases, habitat patches enlarge and coalesce (Figure 1c) as in our assumption depicted in Figure 1b. The PIPL seems to be the shorebird most affected by the changes in its breeding habitat due to sea level rise.
Figure 6Relationships among patch number, size, and connectivity, and fractal dimension of the habitat-specific coastline
Relationships among patch number, size, and connectivity, and fractal dimension of the habitat-specific coastline. 〈s〉vs DK(a), Npvs DK(b), Npvs 〈s〉(c), and 〈c〉vs DK(d) for the threatened, endangered, and at-risk shorebirds (TERs) considered. The dots are the bin averages over 30 simulations for each year for the period 2006-2100. The dashed lines represent the 95% confidence intervals for the dependent variables considered.
2192-1709-1-9-6 The average size and the number of the patches are inversely proportional given the relationship in Figure 6a,b and as shown in Additional file 1: Figure S7. The average patch size 〈s〉 for the shorebirds is not proportional to the average body mass m as possibly expected (Table 1), although the latter scales with the average dispersal length. The 〈s〉is for the PIPL, while it is larger for SNPL and REKN. This emphasizes the controlling role of habitat geomorphology in shaping the patch distribution. The PIPL also depends on the salt-marsh habitat, which t is one of the classes more seriously compromised by SLR. We consider d
l
, the estimated maximum dispersal length, in order to determine the average number of connected patches 〈c〉. d
l
considers rare “Lévy flights” of individuals of the species in the ecosystem. Lévy flights are a special class of random walk with movement displacements drawn from a probability distribution with a power-law tail (the so-called Pareto-Lévy distribution), and they give rise to stochastic processes closely linked to fractal geometry and anomalous diffusion phenomena. Because it has the largest maximum dispersal distance, the REKN has the highest number of connected patches. However, for the three shorebird species 〈c〉increases with the fractal dimension of the patches, indicating a measure of the habitat fragmentation. Because we find that climate change is responsible for the splitting of the patches, rather than their shrinking, and because the dispersal capability of species is not expected to change consistently in the modeled period, the result seems justifiable. The increase in the number of connected patches is explainable because N
p
increases without a drastic reduction in the habitat. The average connectivity of the predicted breeding and wintering patches is an increasing function of the fractal dimension of the patches. The increasing roughness of the Florida coastline due to climate change produces a larger number of patches with smaller dimensions. The increased connectivity would potentially enhance the survivability of the shorebirds despite the decrease in the average size of suitable patches. Thus, the predicted patch patterns for the Florida shorebirds are not the worst case scenario in which both the connectivity and the dimension of the patches are reduced. Further explanation of the land cover, habitat, and patch dynamics is provided in Additional file 1.
Conclusions
Sea level rise due to climate change, beyond being a human-population threat, is shown to strongly affect biodiversity such as residential and migrant shorebird populations in Florida. The integrated patch-prediction modeling framework proposed in this paper constitutes a parsimonious but useful risk assessment tool for species decline with respect to more accurate metapopulation models. In our opinion, the understanding of ecogeomorphological processes at any scale of analysis together with the detection of useful indicators of such dynamics is one of the primary goals to protect biodiversity against the anticipated changes in the landscape due to climate change. On the one hand, it is impossible to consider, or to estimate with low uncertainty, all the factors affecting the processes that govern the distribution of species (e.g., conspecific attractions, interspecific competition, density dependence, sex structure, life history, phenotypic plasticity, and phenological changes in dispersal ability and in breeding/wintering area requirements), the geomorphological processes, and the links and feedbacks among these processes. On the other hand, we believe that a top-down approach of biocomplexity is useful to detect the fundamental drivers of the observed patterns of interest (Schwimmer B64 2008; National Research Council B48 2009; Reinhardt et al. B59 2010). We are aware that many geomorphological and biological processes are not incorporated in the presented model; however, the uncertainty in the quantification of these processes and the interaction of these uncertainties may produce erroneous results in the predictions. The integrated model is capable of providing valuable macroscale predictions with relatively few data and variables. Thus, the model is useful for evaluating conservation actions for increasing the survivability of shorebirds in Florida. We are also confident that the proposed model, properly tuned, can be applied to many different species in coastal ecosystems worldwide that are threatened by sea level rise. We anticipate further development of this model at higher levels of complexity and also for inland sites. The following conclusions are worth mentioning. indent A scale-free distribution of nesting, breeding, and wintering occurrences was detected for the Snowy Plover in Florida. The scale-free distribution was also found for the wintering occurrences of Piping Plover and Red Knot. The distribution was derived through the box-counting technique applied to the breeding and wintering occurrences, which gives a proxy of the fractal dimension of shorebird patches. Empirical evidence shows that the fractal dimension of the occurrences is strongly positively correlated with the coastline fractal dimension, which underlines an ecogeomorphological organization, i.e., a coupling of ecological and geomorphological patterns. The power law held for any season of the shorebird annual cycle, demonstrating the high importance of the physical habitat on species processes.We predicted breeding and wintering patches of shorebirds, simulating land cover (which comprises many coastal wetland types) and habitat suitability at the year scale from 2006 to 2100 as a function of sea level rise. Patches were identified by a set of macroecological criteria, such as area, habitat suitability, and neighboring distance, as a function of the maximum dispersal. The distribution of the predicted patch size was Korčak’s law, whose exponent is half of the fractal dimension of the patches. We validated the model by predicting the observed patch-size distribution and patch patterns from 2002 to 2010 where data were available. We also investigated the perimeter-size relationship for estimating the fractal dimension of the patches at a higher level of complexity because of the calculation of the perimeter. The fractal dimension provided by the perimeter-size relationship provided a median estimate between the values derived from Korčak’s law and the box-counting distribution. Korčak’s law provided the most optimistic scenario of fragmentation in which the probability of finding large patches was the highest, while the box-counting provided the most pessimistic scenario. Hence, the perimeter-area relationship is suggested as the best method to calculate the fractal dimension of the mosaic of habitat patches.The robustness of the Pareto-Lévy distribution of the patch size was verified for predictions of patches from 2006 to 2100. Thus, the scale-invariance of the patch patterns holds in time despite the strong influence of sea level rise. This may be related to a sort of simulated “biological resilience” of species to the external changes (Folke et al. B26 2004) by assuming invariant habitat area and dispersal requirement. Scale-free habitat patterns have proven to be the most resilient to external stressors in previous studies (Kefi et al. 2011). Thus, the shape of the patch-size probability and the fractal dimension when this probability is a power law can be useful indicators to estimate the “degree of stress” of coastal ecosystems. Further research is anticipated to understand when and how the patch-size probability deviates from a Pareto-Lévy behavior. The fragmentation, which is proportional to the fractal dimension of the habitat-specific coastline, varied considerably over time and in particular for the Piping Plover. However, the risk of extirpation in 2100 for SNPL, PIPL, and REKN was not high with respect to 2006. We note that the comparison between final and initial years’ risk should not be the only comparison in evaluating the risk of decline of a species. The overall trend of the fractal dimension in the modeled period has to be evaluated as well.A scaling relationship was found between the fractal dimensions of the patches and of the habitat-specific coastline. The scaling exponent of this relationship appears to be species-independent for the shorebirds considered. Further research is needed to explore the conditions of universality (species- and ecosystem-wise) of this relationship, which may be related to the species considered. The fluctuation in the fractal dimension of the coastline can be assumed to be a valuable ecological indicator for assessing variation in patch patterns of breeding and wintering shorebirds.We demonstrated that habitat loss, fragmentation, and connectivity are three separate concepts. Although these variables are closely linked to each other, their causality is not trivial. For the shorebirds studied, the predicted fragmentation was coupled with habitat loss while the connectivity increased. The fact that the patches, even if smaller, were connected is an extremely positive factor that ensures dispersal and gene flow; thus, the connectivity of patches enhances the survivability of shorebirds. Birth, death, and dispersal processes of a species can overcome the habitat-loss effect and a decrease in the average size of patches. Yet, a lower metapopulation risk of extirpation exists if interpatch migration is allowed (Kindvall and Petersson B37 2000). However, a decrease in the average patch size can potentially increase intra-species competition for foraging (Ritchie B61 1998) and decrease carrying capacity. A possible optimal ecogeomorphological state of the coastal ecosystem may be characterized by the smallest fractal dimension of the coastline that maximizes the compactness of the suitable patches. This configuration also minimizes the fractal dimension of the patches. The highest entropy of this configuration may translate into the smallest energy expenditure of the species that inhabit the habitat, for example, for foraging and breeding activities. The entropy of geomorphological landforms (Nieves et al. B49 2010) may, in fact, be highly correlated with the scale-invariance of ecological patterns such as species-patch patterns.
Endnotes
submitted to Ecological Processes -
Special Issue “Wetlands In a Complex World”, Guest Editor: Dr. Matteo Convertino
Abbreviations
SNPL: Snowy Plover; PIPL: Piping Plover; REKN: Red Knot; TER: threatened, endangered, and at risk; SLAMM: Sea Level Affecting Marshes Model; SLR: sea level rise; Df: fractal dimension of the coastline (from box-counting); Db: fractal dimension of the breeding and wintering occurrences (from box-counting); DK: fractal dimension of the patches (from Korčak’s law); Dc: fractal dimension of the patches (from perimeter-size relationship); S: patch-size; p: patch perimeter; P(hs): habitat suitability score; SI: suitability index; Sp: minimum population patch-size; Sb/w: minimum breeding/wintering patch-size; hr: home-range; hrd: home-range distance; dl: maximum dispersal length.
Competing interests
The authors declare that they have no competing interests.
Author’s contributions
MC designed the study, managed and analyzed the data, wrote the model (box-counting and patch delineation model), developed the theory, and wrote the manuscript. AB assisted in making the calculations and analysis, and helped in writing the manuscript. GAK and RMC participated in the habitat suitability modeling framework and reviewed the manuscript. IL supervised the whole work, and reviewed the manuscript by providing a practical angle to this research for effective environmental management. All authors read and approved the final manuscript.
Authors’ information
MC is Research Scientist at the University of Florida, Gainesville, and a Contractor of the Engineering Research and Development Center of the US Army Corps of Engineers at the Risk and Decision Science Team. AB is currently a financial analyst at Frontier Airlines. AB got his B.Sc and M.Sc. from MIT, Civil and Environmental Engineering program. AB performed his research internship at the Risk and Decision Science Team in the summer of 2011. GAK and RMC are Associate and Professor at the University of Florida, Gainesville, respectively. IL is team leader of the Risk and Decision Science Team of the Engineering Research and Development Center of the US Army Corps of Engineers.
bm
ack
Acknowledgements
This research was supported by the US Department of Defense, through the Strategic Environmental Research and Development Program (SERDP), Project SI-1699. M.C. acknowledges the funding of project “Decision and Risk Analysis Applications Environmental Assessment and Supply Chain Risks” for his research at the Risk and Decision Science Team. The computational resources of the University of Florida High-Performance Computing Center (http://hpc.ufl.edu) are kindly acknowledged. The authors cordially thank Dr. RA Fisher (Engineering Research and Development Center of the US Army Corps of Engineers) and the Eglin Air Force Base personnel for their help in obtaining the data and for the useful information about the breeding information of SNPL. Tyndall Air Force Base and Florida Wildlife Commission are also gratefully acknowledged for the assistance with the data. We thank M.L. Chu-Agor (currently at the Center of Environmental Sciences, Department of Biology and Earth and Atmospheric Sciences, Saint Louis University, St. Louis, MO) for her computational effort with SLAMM at the University of Florida. Permission was granted by the USACE Chief of Engineers to publish this material. The views and opinions expressed in this paper are those of the individual authors and not those of the US Army or other sponsor organizations.
refgrp The impact of sea level rise on snowy plovers in Florida: integrating geomorphological, habitat, and metapopulation modelsAiello-LammensMEChu-AgorMLConvertinoMFischerRALinkovIResit AkcakayaHGlobal Change Biol2011173644lpage 365410.1111/j.1365-2486.2011.02497.xEffects of patch shape in intertidal algal mosaics: roles of area, perimeter and distance from edgeAiroldiLMar Biol200314363965010.1007/s00227-003-1119-3Florida Panhandle Shorebird Working GroupAllianceFS2010Tech. rep FWC/Audubon Florida/USFWS. http://www.flshorebirdalliance.org/resources-pages/maps.htmlcnm AudubonAmerica’s top ten most endangered birds. Tech. reppublisher National Audubon Society, New york2006pmpid 23359901BakPHow nature works: the science of self-organized criticalitySpringer-Verlag, Berlin1999A numerical retro-action model relates rocky coast erosion to percolation theoryBaldassarriASapovalBFelixS2012Arxiv. http://arxiv.org/pdf/1202.4286v1.pdfScaling, optimality and landscape evolutionBanavarJRColaioriFFlamminiAMaritanARinaldoAJ Stat Phys200110413310.1023/A:1010397325029BattyMFractal citiesAcademic Press, San Diegoeditor Longley, P1994Scaling in animal group-size distributionsBonabeauEDagornLFreonPProc Nat Acad Sci19999644724477pubidlist 10.1073/pnas.96.8.4472pmcid 16356link fulltext 10200286Hierarchy in social organizationBuldyrevSDokholyanNErramilliSHongMKimJMalescioGStanleyHPhys Stat Mech Appl20033303-465365910.1016/j.physa.2003.09.041Florida beach-nesting bird report, 2005-2008BurneyC2009Tech. rep. Florida Fish and Wildlife Conservation Commission, Tallahassee. http://www.flshorebirdalliance.org/pdf/2005-2008_FWC_BNB_Report.pdfCoastal change analysis program regional land coverC-CAP2009Tech. rep. NOAA, Washington DC. http://www.csc.noaa.gov/digitalcoast/data/ccapregional/Exploring sea level rise vulnerability of coastal habitats using global sensitivity and uncertainty analysisChu-AgorMMuñoz-CarpenaRKikerGEmanuelssonALinkovIEnviron Model Software201126559360410.1016/j.envsoft.2010.12.003Application of SLAMM 4.1 to nine sites in FloridaCloughJ2006Warren Pinnacle Consulting, Waitsfield, VT. http://warrenpinnacle.com/prof/SLAMM/NWF_SLAMM_FLORIDA_2-16-2006.docThe Sea Level Affecting Marshes ModelCloughJS2010Tech. rep. Warren Pinnacle Consulting, Waitsfield, VT. http://warrenpinnacle.com/prof/SLAMM6/SLAMM6_Technical_Documentation.pdfIntegrated modeling to mitigate climate change risk due to sea-level rise of imperiled shorebirds on Florida coastal military installations.ConvertinoMKikerGChu-AgorMMuñoz-CarpenaRMartinezCAiello-LammensMAkçakayaHFisherRLinkovISpringer, Dordrecht2010Anthropogenic renourishment feedback on shorebirds: a multispecies Bayesian perspectiveConvertinoMDonoghueJChu-AgorMKikerGMunoz-CarpenaRFischerRLinkovIEcol Eng2011a371184119410.1016/j.ecoleng.2011.02.019Scale- and resolution-invariance of suitable geographic range for shorebird metapopulationsConvertinoMKikerGMunoz-CarpenaRChu-AgorMFischerRLinkovIEcol Complexity2011b8436437610.1016/j.ecocom.2011.07.007Do tropical cyclones shape shorebird habitat patterns? Biogeoclimatology of Snowy Plovers in FloridaConvertinoMElsnerJMuñoz-CarpenaRKikerGFisherRLinkovIPLoS ONE2011c61p1From river basins to elephants to bacteria colonies: aggregate-size spectrum of animate and inanimate speciesConvertinoMSiminiFCataniFKikerGPLoS ONE (in press)2012Integrating multiple modelling approaches to predict the potential impacts of climate change on species’ distributions in contrasting regions: comparison and implications for policydel BarrioGHarrisonPBerryPButtNSanjuanMPearsonRDawsonTEnviron Sci Policy20069212914710.1016/j.envsci.2005.11.005The evolution of normsEhrlichPRLevinSAPLoS Biol200536e19410.1371/journal.pbio.0030194114949115941355Piping plover (charadrius melodus).Elliott SmithEHaigSCornell Lab of Ornithology, Ithaca200410.2173/bna.2F-DEPState Geological Map FDEP-GEO2001Tech. rep., Florida Department of Environmental Protection (data from, FGDL, University of Florida GEOPLAN Center). http://www.dep.state.fl.us/geology/gisdatamaps/state_geo_map.htmPetition to list the red knot (caladris canutus rufa) as endangered and request for emergency listing under the endangered species actFallonF2005Tech. rep. Delaware Riverkeeper Network, American Littoral Society, Delmarva Ornithological Society, Delaware Chapter of the Sierra Club, New Jersey Audubon Society. http://www.fws.gov/northeast/redknot/riverkeeper.pdfRegime shifts, resilience and biodiversity in ecosystem managementFolkeCCarpenterSWalkerBSchefferMElmqvistTGundersonLHollingCAnnu Rev Ecol Evol Syst20043555758110.1146/annurev.ecolsys.35.021103.105711Florida beach-nesting birds websiteFWC2010Tech. rep. Florida Fish and Wildlife Conservation Commission. http://legacy.myfwc.com/bnb/Red knot (calidris canutus).HarringtonBCornell Lab of Ornithology, Ithaca200110.2173/bna.563Ecol AppHelzerCJelinskiD1999914481458Status and Distribution of Snowy Plover in Florida, 2006HimesJDouglassNPrunerRCroftASeckingerE2006Tech. rep. Florida Wildlife Conservation Commission. http://www.flshorebirdalliance.org/pdf/Himes_Douglass-2006_SNPL_Report.pdfEnvironmental context explains Lévy and Brownian movement patterns of marine predatorsHumphriesNetal Nature201046573011066106910.1038/nature0911620531470The fractal dimension as a measure of the quality of habitatsImreARBogaertJActa Biotheoretica200452415614963403Fractal bird nest distribution produces scale-free colony sizesJovaniRTellaJProc R Soc B20072742465246910.1098/rspb.2007.0527227497017666378Truncated power laws reveal a link between low-level behavioral processes and grouping patterns in a colonial birdJovaniRSerranoDTellaJLAdlerFRUrsúaEPLoS ONE200831992+10.1371/journal.pone.0001992Spatial vegetation patterns and imminent desertification in Mediterranean arid ecosystemsKéfiSRietkerkMAladosCLPueyoYPapanastasisVPElaichAde RuiterPCNature200744921321710.1038/nature0611117851524Robust scaling in ecosystems and the meltdown of patch size distributions before extinctionKefiSRietkerkMRoyMFrancAde RuiterPPascualMEcol Lett2011141293510.1111/j.1461-0248.2010.01553.x21070563onsequences of modelling interpatch migration as a function of patch geometry when predicting metapopulation extinction riskKindvallOPeterssonAEcol Modell2000129110110910.1016/S0304-3800(00)00215-5A coastal land cover classification system for the noaa coastwatch change analysis projectKlemasVDobsonJFergusonRHaddadKJ Coastal Res199393862872Deux types fondamentaux de distribution staffstiqueKorcakJBull Inst Int Star194030295307Status and Distribution of Snowy Plover in Florida, 2002LamonteKDouglassN2002Tech. rep. Florida Wildlife Conservation Commission. http://www.flshorebirdalliance.org/pdf/Lamonte_Douglass-2002_SNPL_Report.pdfSmall home ranges and high site fidelity in red knots (calidris c. canutus) wintering on the banc dArguin, mauritaniaLeyrerJSpaansBCamaraMPiersmaTJ Ornithology2006147237638410.1007/s10336-005-0030-8Corridordesigner: Arcgis tools for designing and evaluating corridorsMajkaDJennessJBeierPCorridorDesigner2007
http://corridordesign.org
How long is the coast of britain? Statistical self-similarity and fractional dimensionMandelbrotBScience1967156377510.1126/science.156.3775.636MandelbrotBThe Fractal Geometry of NatureW.H. Freeman, New York1982Application of a barrier island translation model to the millennial-scale evolution of Sand Key, FloridaMasettiRFagherazziSMontanariAContinental Shelf Res2008281116112610.1016/j.csr.2008.02.021Fractality of eroded coastlines of correlated landscapesMoraisPAOliveiraEAAraújoNAMHerrmannHJAndradeJSPhys Rev E201184016,10210.1103/PhysRevE.84.016102Model relevance: frameworks for, exploring the complexity-sensitivity-uncertainty trilemma.MullerSMuñoz-CarpenaRKikerGNATO book, Amsterdam2010National Research Council, NAS (ed)A New Biology for the 21st CenturyThe National Academies Press, Washington, DC2009Maximum entropy distributions of scale-invariant processesNievesVWangJBrasRWoodEPhys Rev Lett201010511118,701+10.1103/PhysRevLett.105.118701Scaling properties in landscape patterns: New Zealand ExperienceNikoraVPearsonCShankarULandscape Ecol199914173310.1023/A:1008077914643Snowy plover (charadrius alexandrinus).PageGStenzelLWarrinerJPatonP2009In: Cornell I Lab of Ornithology, Poole A (eds) The Birds of North America Online, vol. 5. http://bna.birds.cornell.edu/bna/species/15410.2173/bna.154Environmental dynamics: Simplicity versus complexityPaolaCLeederMNature2011469383910.1038/469038a21209652Cluster-size distributions: signatures of self-organization in spatial ecologiesPascualMManojitRGuichardFFlierlGPhil Trans R Soc Lond B200235765766610.1098/rstb.2001.0983Simple models for complex systems: exploiting the relationship between local and global densitiesPascualMRoyMLaneriKTheor Ecol20114221122210.1007/s12080-011-0116-2Factors affecting interannual movements of snowy ploversPatonPEdwardsTThe Auk1996113353454310.2307/4088973Maximum entropy modeling of species geographic distributionsPhillipsSAndersonRSchapireREcol Modell20061903-423125910.1016/j.ecolmodel.2005.03.026Modeling of species distributions with maxent: new extensions and a comprehensive evaluationPhillipsSJMiroslavDEcography200831216117510.1111/j.0906-7590.2008.5203.xAssessing habitat selection, reproductive performance, and the affects of anthropogenic disturbance of the Snowy Plover along the Florida Gulf coastPrunerR2010Master’s thesis, University of Florida, Gainesville, USADynamic interactions of life and its landscape: feedbacks at the interface of geomorphology and ecologyReinhardtLJerolmackDCardinaleBJVanackerVWrightJEarth Surf Processes Landforms2010357810110.1002/esp.1912Geomorphological signatures of varying climateRinaldoADietrichWERigonRVogelGKRodrlguez-LturbeINature199537463263510.1038/374632a0Scale-dependent foraging and patch choice in fractal environmentsRitchieMEvolutionary Ecol19981230933010.1023/A:1006552200746Positive feedbacks promote power-law clustering of Kalahari vegetationScanlonTMCaylorKKLevinSARodriguez-IturbeINature200744920921210.1038/nature0606017851523SchneiderDTellaJScaling theory: application to marine ornithologyEcosystems2002573674810.1007/s10021-002-0156-yA temporal geometric analysis of eroding marsh shorelines: can fractal dimensions be related to processesSchwimmerRAJ Coastal Res20081152158Effect of sea level rise on piping plover (charadrius melodus) breeding habitatSeaveyJRGilmerBMcGarigalKMBiol Conserv201010.1016/j.biocon.2010.09.017SoléRSelf-organization in Complex EcosystemsPrinceton University Press, Princeton, NJBascompte, J2006Long-distance breeding dispersal of Snowy Plovers in western North AmericaStenzelLWarrinerJWarrinerJWilsonKBidstrupFPageGJ Animal Ecol19946388790210.2307/5266Survival and natal dispersal of juvenile snowy plovers (Charadrius alexandrinus) in central coastal CaliforniaStenzelLPageGWarrinerJWarrinerJGeorgeDEysterCRamerBNeumanKThe Auk200712431023103610.1642/0004-8038(2007)124[1023:SANDOJ]2.0.CO;2Piping Plover (Charadrius melodus) 5-Year Review: Summary and EvaluationUSFWS2009Tech. rep., US Fish and Wildlife Service. http://www.fws.gov/northeast/endangered/PDF/Piping_Plover_five_year_review_and_summary.pdfNational Elevation DatasetUSGS2010Tech. rep., United State Geological Survey. http://ned.usgs.gov/Data from the 2006, International Piping Plover Census (IPIPLC),USGS-FWS2009Tech. rep., United State Geological Survey and USFWS. http://www.flshorebirdalliance.org/pdf/Elliot-Smith_Haig-2006_PIPL_Report.pdfMating system and reproductive success of a small population of polygamous Snowy PloversWarrinerJSWarrinerJCPageGWStenzelLEWilson Bull1986981537


xml version 1.0 encoding utf-8 standalone no
mets ID sort-mets_mets OBJID sword-mets LABEL DSpace SWORD Item PROFILE METS SIP Profile xmlns http:www.loc.govMETS
xmlns:xlink http:www.w3.org1999xlink xmlns:xsi http:www.w3.org2001XMLSchema-instance
xsi:schemaLocation http:www.loc.govstandardsmetsmets.xsd
metsHdr CREATEDATE 2013-01-30T12:07:36
agent ROLE CUSTODIAN TYPE ORGANIZATION
name BioMed Central
dmdSec sword-mets-dmd-1 GROUPID sword-mets-dmd-1_group-1
mdWrap SWAP Metadata MDTYPE OTHER OTHERMDTYPE EPDCX MIMETYPE textxml
xmlData
epdcx:descriptionSet xmlns:epdcx http:purl.orgeprintepdcx2006-11-16 xmlns:MIOJAVI
http:purl.orgeprintepdcxxsd2006-11-16epdcx.xsd
epdcx:description epdcx:resourceId sword-mets-epdcx-1
epdcx:statement epdcx:propertyURI http:purl.orgdcelements1.1type epdcx:valueURI http:purl.orgeprintentityTypeScholarlyWork
http:purl.orgdcelements1.1title
epdcx:valueString Shorebird patches as fingerprints of fractal coastline fluctuations due to climate change
http:purl.orgdctermsabstract
Abstract
Introduction
The Florida coast is one of the most species-rich ecosystems in the world. This paper focuses on the sensitivity of the habitat of threatened and endangered shorebirds to sea level rise induced by climate change, and on the relationship of the habitat with the coastline evolution. We consider the resident Snowy Plover (Charadrius alexandrinus nivosus), and the migrant Piping Plover (Charadrius melodus) and Red Knot (Calidris canutus) along the Gulf Coast of Mexico in Florida.
Methods
We analyze and model the coupled dynamics of habitat patches of these imperiled shorebirds and of the shoreline geomorphology dictated by land cover change with consideration of the coastal wetlands. The land cover is modeled from 2006 to 2100 as a function of the A1B sea level rise scenario rescaled to 2 m. Using a maximum-entropy habitat suitability model and a set of macroecological criteria we delineate breeding and wintering patches for each year simulated.
Results
Evidence of coupled ecogeomorphological dynamics was found by considering the fractal dimension of shorebird occurrence patterns and of the coastline. A scaling relationship between the fractal dimensions of the species patches and of the coastline was detected. The predicted power law of the patch size emerged from scale-free habitat patterns and was validated against 9 years of observations. We predict an overall 16% loss of the coastal landforms from inundation. Despite the changes in the coastline that cause habitat loss, fragmentation, and variations of patch connectivity, shorebirds self-organize by preserving a power-law distribution of the patch size in time. Yet, the probability of finding large patches is predicted to be smaller in 2100 than in 2006. The Piping Plover showed the highest fluctuation in the patch fractal dimension; thus, it is the species at greatest risk of decline.
Conclusions
We propose a parsimonious modeling framework to capture macroscale ecogeomorphological patterns of coastal ecosystems. Our results suggest the potential use of the fractal dimension of a coastline as a fingerprint of climatic change effects on shoreline-dependent species. Thus, the fractal dimension is a potential metric to aid decision-makers in conservation interventions of species subjected to sea level rise or other anthropic stressors that affect their coastline habitat.
http:purl.orgdcelements1.1creator
Convertino, Matteo
Bockelie, Adam
Kiker, Gregory A
Muñoz-Carpena, Rafael
Linkov, Igor
http:purl.orgeprinttermsisExpressedAs epdcx:valueRef sword-mets-expr-1
http:purl.orgeprintentityTypeExpression
http:purl.orgdcelements1.1language epdcx:vesURI http:purl.orgdctermsRFC3066
en
http:purl.orgeprinttermsType
http:purl.orgeprinttypeJournalArticle
http:purl.orgdctermsavailable
epdcx:sesURI http:purl.orgdctermsW3CDTF 2012-10-30
http:purl.orgdcelements1.1publisher
Springer
http:purl.orgeprinttermsstatus http:purl.orgeprinttermsStatus
http:purl.orgeprintstatusPeerReviewed
http:purl.orgeprinttermscopyrightHolder
Matteo Convertino et al.; licensee BioMed Central Ltd.
http:purl.orgdctermslicense
http://creativecommons.org/licenses/by/2.0
http:purl.orgdctermsaccessRights http:purl.orgeprinttermsAccessRights
http:purl.orgeprintaccessRightsOpenAccess
http:purl.orgeprinttermsbibliographicCitation
Ecological Processes. 2012 Oct 30;1(1):9
http:purl.orgdcelements1.1identifier
http:purl.orgdctermsURI http://dx.doi.org/10.1186/2192-1709-1-9
fileSec
fileGrp sword-mets-fgrp-1 USE CONTENT
file sword-mets-fgid-0 sword-mets-file-1
FLocat LOCTYPE URL xlink:href 2192-1709-1-9.xml
sword-mets-fgid-1 sword-mets-file-2 applicationpdf
2192-1709-1-9.pdf
sword-mets-fgid-3 sword-mets-file-3
2192-1709-1-9-S1.PDF
sword-mets-fgid-4 sword-mets-file-4 videox-msvideo
2192-1709-1-9-S2.AVI
structMap sword-mets-struct-1 structure LOGICAL
div sword-mets-div-1 DMDID Object
sword-mets-div-2 File
fptr FILEID
sword-mets-div-3
sword-mets-div-4
sword-mets-div-5