Identification of a spatially efficient portfolio of priority conservation sites in marine and estuarine areas of Florida

AQUATIC CONSERVATION: MARINE AND FRESHWATER ECOSYSTEMS

Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

Published online in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/aqc.992

Identification of a spatially efficient portfolio of priorityconservation sites in marine and estuarine areas of Florida

LAURA GESELBRACHTa,�, ROBERTO TORRESa, GRAEME S. CUMMINGb,y, DANIEL DORFMANc,z

MICHAEL BECKc and DOUGLAS SHAWa

aThe Nature Conservancy, 222 S. Westmonte Drive, Suite 300, Altamonte Springs, Florida 32714 USAbDepartment of Wildlife Ecology and Conservation, University of Florida, Gainesville, Florida 32611

cThe Nature Conservancy, University of California Santa Cruz, Center for Ocean Health, Santa Cruz, California 95060 USA

ABSTRACT

1. A systematic conservation planning approach using benthic habitat and imperilled species data along withthe site prioritization algorithm, MARXAN, was used to identify a spatially efficient portfolio of marine andestuarine sites around Florida with high biodiversity value.2. Ensuring the persistence of an adequate geographic representation of conservation targets in a particular

area is a key goal of conservation. In this context, development and testing of different approaches to spatially-explicit marine conservation planning remains an important priority.3. This detailed case study serves as a test of existing approaches while also demonstrating some novel ways in

which current methods can be tailored to fit the complexities of marine planning.4. The paper reports on investigations of the influence of varying several algorithm inputs on resulting portfolio

scenarios including the conservation targets (species observations, habitat distribution, etc.) included,conservation target goals, and socio-economic factors.5. This study concluded that engaging stakeholders in the development of a site prioritization framework is a

valuable strategy for identifying broadly accepted selection criteria; universal target representation approachesare more expedient to use as algorithm inputs, but may fall short in capturing the impact of historic exploitationpatterns for some conservation targets; socio-economic factors are best considered subsequent to theidentification of priority conservation sites when biodiversity value is the primary driver of site selection; andthe influence of surrogate targets on portfolio selection should be thoroughly investigated to ensure unintendedeffects are avoided.6. The priority sites identified in this analysis can be used to guide allocation of limited conservation and

management resources.Copyright r 2008 John Wiley & Sons, Ltd.

Received 28 November 2007; Revised 16 April 2008; Accepted 26 May 2008

KEY WORDS: systematic conservation planning; geographic representation; resilience; marine reserve

INTRODUCTION

Conservation efforts to protect biodiversity and maintain the

integrity of ecological systems are limited by time, funding and

staff resources. Often, conservation need is urgent. In the past,

decisions regarding the selection of priority marine

conservation areas have frequently been made on an ad hoc,

as needed, or opportunity driven basis (Roberts et al., 2003a).

While the designation of many managed areas has been

valuable for conservation goals, some of the most ecologically

important areas in a particular planning region may have been

missed and some resources may be under-represented. A range

*Correspondence to: Laura Geselbracht, The Nature Conservancy, 222 S. Westmonte Drive, Suite 300, Altamonte Springs, Florida 32714 USA.E-mail: [email protected] Address: Percy FitzPatrick Institute, DST/NRF Center of Excellence, University of Cape Town, Cape Town, South Africa.zPresent Address: Intelligent Marine Planning, St Petersburg, Florida 33704 USA.

Copyright r 2008 John Wiley & Sons, Ltd.

of more systematic approaches, which may include the use of

site selection algorithms, have been pioneered in recent times

to facilitate the identification of priority marine and estuarine

sites (Pressey et al., 1994; Ball and Possingham, 2000;

Possingham et al., 2000; Salm et al., 2000; Beck, 2003;

Groves, 2003; Leslie, 2005). Despite their relevance, many of

these tools have had little application in real-world marine

conservation planning. Consequently there is a clear need for

further case studies that explore the strengths and weaknesses

of different systematic approaches and develop novel ways of

applying these methods in cases where conventional

applications are difficult or inappropriate. This article seeks

to contribute to the further development of systematic marine

conservation planning efforts through a detailed exploration

of their application in the state of Florida, USA.

Systematic conservation approaches generally require

identifying the best sites for concentrating conservation and

management activities in a particular planning region. Site

selection may be influenced by how much of the conservation

targets are represented at the sites, the redundancy of the

occurrences (to spread the risk in the face of stress), viability of

the resource occurrences represented, and connectivity between

occurrences to ensure replenishment and genetic diversity

(Shaffer and Stein, 2000; Groves, 2003). Economic and social

criteria may also be applied to either select or avoid areas of high

importance (Hockey and Branch, 1997; Roberts et al., 2003b).

Which selection criteria to use and how their use is applied will be

driven by the specific goals of each planning effort.

Site selection algorithms, such as MARXAN, have been

developed to accept multiple selection criteria as inputs and

have been used to identify priority conservation areas (e.g.

marine reserve networks) for numerous seascapes over the last

several years (Beck and Odaya, 2001; Floberg et al., 2004;

DeBlieu et al., 2005). The basic aim of these optimization

algorithms is to achieve conservation target goals in a spatially

efficient manner. Such algorithms allow for the evaluation of

dozens of conservation targets and hundreds of possible

conservation sites in a practically limitless arrangement. In

the past, such efforts have relied solely on expert opinion to

identify priority sites, but the vast number of configurations to

evaluate makes this a daunting task (Leslie et al., 2003). The

use of site selection algorithms to identify priority conservation

sites does not replace the need to engage resource experts, but

rather simplifies the task and provides a tool for explicit

identification of selection criteria and trade-offs.

In areas such as Florida where an extensive system of existing

marine and estuarine managed sites already exists, systematic

conservation planning using site selection algorithms can be used

to investigate the adequacy of the existing management regime

and provide a means for prioritizing the allocation of conservation

and management resources (funding and staff) around the

planning area. This study explored the influence of a number of

selection criteria on the development of a portfolio of priority

marine and estuarine conservation sites based on spatially efficient

representation of conservation features. It utilized existing biotic

resources and socio-economic datasets along with the site selection

algorithm, MARXAN, to identify several alternative portfolios of

priority sites. The results were vetted through a series of expert

workshops and utilized the collective advice of reviewers to

develop a preferred portfolio. The determination of the most

appropriate conservation actions to take at any particular site is

left to more detailed site planning processes.

METHODS

MARXAN and existing geospatial datasets of marine and

estuarine resources were utilized to develop several potential

conservation portfolios for the marine and estuarine areas

surrounding Florida. To accomplish this, planning area

boundaries and subregions, conservation targets, appropriate

target distribution and socio-economic use datasets, and

alternative approaches for setting representation goals were

identified. An index for spatially representing socio-economic

activities likely to have an ‘irreversible’ adverse impact on

biodiversity and/or resource viability was also created.

Evaluation of the efficiency of the portfolio scenarios in

terms of spatial representation and attainment of conservation

target goals was carried out. Expert review workshops were

held throughout the priority site identification process, and the

comments from these were utilized, including those on a

number of alternative portfolios developed to help shape

selection of a preferred alternative.

Planning area and geographic stratification

This study encompassed marine and estuarine habitats

surrounding the state of Florida in the USA. The 1350 mile

Florida coastline supports a diverse and productive

assemblage of marine and estuarine systems. Owing to its

position (i.e. jutting from the North American continent south

towards the Caribbean Sea), Florida’s marine systems range

from temperate in the north to tropical in the south. Extensive

salt marsh and mangrove systems grace Florida’s low energy

coastlines. Beach and barrier island complexes are found in

higher energy areas. Vast, globally significant seagrass

meadows are present along portions of the Gulf Coast and

in Florida Bay. Large expanses of mangrove forests dominate

coastal areas of South Florida, which is also home to the third

largest barrier reef tract in the world. Large and small estuaries

and coastal rivers occur all along the Florida coast. These

highly productive systems support key life stages of many

commercially and recreationally important species such as

pink shrimp, stone crab, lobster, scallops, clams, groupers,

snappers, mullet and numerous other species of game fish.

The Florida planning area for the project was broadly

defined by setting the offshore boundary at the 500m isobath

which comfortably encompasses the continental margin. The

inland boundary was defined by the zone of saltwater influence

as identified by the National Wetlands Inventory marine and

estuarine classifications (USFWS, 1979). The total area

described by these boundaries is approximately 28.4 million

hectares. To capture biogeophysical differences over this large

planning area, it was divided into eight subregions based

primarily on coastal geomorphology and faunal assemblages

(Davis, 1997; Randazzo and Halley, 1997; Lodge, 2005)

(Figure 1).

Conservation targets and data sources

For the priority areas analysis, benthic habitats and vulnerable

species were selected as conservation targets to represent

biodiversity. A hierarchical marine and estuarine habitat

classification scheme developed by the State of Florida

(System for Classification of Habitats in Estuarine and

Marine Environments) was used to categorize the habitat

types used in this analysis (Madley et al., 2002). The most

L. GESELBRACHT ET AL.

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

recent geospatial datasets available were assembled for each

broad marine and estuarine benthic habitat category. Selected

datasets were limited to those that covered at minimum an

entire subregion. Datasets were primarily obtained from state,

federal and local government agencies and mostly covered only

areas within state waters. In three cases datasets were derived.

The coastal tidal river or stream dataset was derived from a

data set representing the inland limit of tide available through

the State of Florida and the National Hydrography Dataset.

The fish spawning aggregation dataset was derived from

interviewing fishermen and comparing with similar data

collected by Environmental Defense (K. Lindeman, personal

communication, 2005). The third derived dataset, benthic

complexity, was created from bathymetry data. It was utilized

to provide some means for predicting ecological importance

where little or no geospatial data were available on benthic

habitats in the planning area. Benthic complexity was the only

dataset used in the study that covered almost the entire study

area. Benthic complexity was estimated using an ArcInfo GIS

model developed by Duke University Marine Geospatial

Ecology Laboratory (2005). The model used bathymetry

data (90m grid-scale resolution) and four geophysical

features: depth, topographic variety, amplitude of

topographic change and substrate type. The model is based

on the strong correlation between benthic complexity and

species richness (Ardron, 2002). Topographic variety was

classified as flat, slope, ridge and canyon. Sediment classes

were extrapolated from data in the Atlantic and Gulf States

Marine Fisheries Commissions’ Southeast Area Monitoring

and Assessment Program (SEAMAP) and the USGS

usSEABED Project. Application of the resulting model

identified areas of relatively more complex bottom

topography. Benthic complexity could not be calculated for

a few areas where bathymetry data were unavailable, including

some shallow areas (o50m) of the Big Bend Subregion and

the Florida Bay/Ten Thousand Islands area.

Ecologically vulnerable species targets and/or aggregations

were also included in the site selection process to ensure the

inclusion of ecologically important sites that might not

otherwise be identified from benthic habitat information

alone. The number of species targets included was limited to

ensure balance with the habitat targets, and to avoid a

situation in which species targets alone drove algorithm

results. The number of species targets included in this

analysis was limited by using the objective selection criteria

identified below:

� Globally, regionally or state imperilled species (G1-G2/

G3), S1–S3, State Species of Special Concern—SSC),

Figure 1. Planning area and subregions designated for this study. Outer planning area boundary is the 500m isobath. Inland boundary is the zone ofsaltwater influence as identified by the National Wetlands Inventory (USFWS, 1979). Subregions are named as follows: 1—Northeast Florida, 2—East-Central Florida, 3—Southeast Florida, 4—Florida Keys/Florida Bay, 5—Southwest Florida/Ten Thousand Islands, 6—West-Central

Florida, 7—Big Bend, and 8—Northwest Florida/Panhandle.

IDENTIFICATION OF PRIORITY CONSERVATION SITES IN MARINE AREAS OF FLORIDA

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

IUCN red-listed, federally listed/candidate species and

American Fisheries Society threatened or endangered

distinct population segments; and

� Vulnerable breeding aggregations subject to harvest or

other forms of direct human disturbance (e.g. marine

dependent bird nesting aggregations and known fish

spawning aggregations).

In all, 12 benthic habitat and 35 species/aggregation targets

were included in the study. Species targets were only included

in subregions where they are known to occur, and where data

on their distribution were uniformly collected for most of the

subregion, or data had been collected over a long enough

period that the discovery of a significant number of additional

occurrence sites is not anticipated. The set of conservation

targets used in the analysis is listed in Table 1. Most of the

conservation targets that were used in this study only covered

areas within state waters (3 nautical miles on the Atlantic

Coast and 9 nautical miles on the Gulf of Mexico coast) with

the exception of benthic complexity, marine hardbottom, in-

water sea turtles and the well-documented deep Oculina banks

(Reed, 2004).

Selection of priority sites

The optimization algorithm, MARXAN was used to aid in

identification of priority sites. MARXAN was developed by

Hugh Possingham and Ian Ball of the University of

Queensland, Australia, and along with related site selection

algorithms has been used for a variety of marine applications

(Ball and Possingham, 2000; Possingham et al., 2000).

MARXAN enabled the specified conservation target goals to

be achieved while minimizing the total area of all sites selected.

While biodiversity and ecosystem conservation objectives may

direct us towards maximizing the size and number of

conservation areas, practical considerations of funding and

staffing levels lead towards the identification of focal areas in

which to concentrate efforts. MARXAN seeks to minimize

total cost of the selected sites using the following objective

function:

Total Cost ¼X

i

Cost site i þX

j

Penalty cost for element j

þ wb

Xboundary length

Cost may be represented as total area of the sites, but may also

include socio-economic costs. Socio-economic costs may be

thought of in terms of existing human activities or structures

that may render conservation more expensive. Penalty cost,

also known as the species penalty factor, represents the penalty

given for not adequately representing a conservation feature

and is summed for all conservation features. Boundary length

is the total perimeter of all the sites. Its inclusion in the

objective function controls fragmentation. Fragmentation of

the solution can be reduced (i.e. clumpiness of the solution

increased) by setting the constant Wb to a value greater than 0.

If the constant is given a value of 0, then boundary length has

no effect on the solution. Three optimization methods are used

for the objective function: the iterative improvement

Table 1. Conservation targets represented in site prioritization framework

Benthic habitat Species and vulnerable aggregations

Coral reef (incl. deep Oculina Banks) Florida manatee, Trichechus manatus latirostrisMangrove forest Northern right whale, Eubalaena glacialisBeach/surf zone American oystercatcher, Haematopus palliatusSalt marsh Black skimmer, Rynchops nigerSubmerged aquatic vegetation Brown pelican, Pelecanus occidentalisCoastal tidal river or stream Least tern, Sterna antillarumTide flats Piping plover, Charadrius melodusMarine hardbottom Reddish egret, Egretta rufescensBivalve reef (Oyster) Roseate spoonbill, Ajaia ajaiaAnnelid worm reef (Sabellariidae) Roseate tern, Sterna dougalliiOcean inlets and passes Snowy egret, Egretta thulaBenthic complexity Snowy plover, Charadrius alexandrinus tenuirostris

Waterbird nesting sitesAmerican crocodile, Crocodylus acutusSea turtle nesting sitesSea turtles, in-water surveysDiamondback terrapin (Ornate, Mississippi and Carolina), Malaclemys terrapinmacrospilota, M.t. pileata and M.t. centrataSmalltooth sawfish, Pristis pectinataSlashcheek goby, Ctenogobius pseudofasciatusRiver goby, Awaous bananaBigmouth sleeper, Gobiomorus dormitorMangrove Rivulus, Rivulus marmoratusOpossum pipefish, Microphis brachyurus lineatusStriped croaker, Bairdiella sanctaeluciaeGulf sturgeon, Acipenser oxyrhynchus desotoiAtlantic sturgeon, Acipenser oxyrhynchus oxyrhynchusShortnose sturgeon, Acipenser brevirostrumSaltmarsh topminnow, Fundulus jenkinsiAlabama shad, Alosa alabamaeKey silverside, Menidia conchorumElkhorn coral, Acropora palmataJohnson’s seagrass, Halophila johnsoniiSpawning aggregations, harvested species

L. GESELBRACHT ET AL.

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

algorithm, simulated annealing algorithm and greedy heuristic.

MARXAN starts with an initial random solution, compares

sequential random changes to the original solution and selects

the better of the two for the next iteration. As this iterative

process continues, MARXAN becomes increasingly selective

as it considers improved alternative solutions. A full

description of MARXAN can be found at the University of

Queensland, Ecology Centre website (www.ecology.uq.edu.au/

marxan.htm).

The parameter values selected for this analysis and the

rationale for doing so are described below:

Planning units

The planning area was divided into 1500 ha hexagonal

planning units resulting in a total of 18 943 units. The

1500 ha planning unit size was selected to provide fine

enough detail to resolve habitat within coastal features

(especially within bays and estuaries) for statewide analysis

while not overwhelming processing capabilities or exceeding

the resolution of the habitat data.

Conservation target goals

The amount of each target required to meet conservation goals

is a key planning parameter and input to MARXAN. In

priority conservation site identification, the purpose may be to

identify the minimum amount of each habitat type required to

maintain fully viable populations and communities into the

future. So far, there is no universal agreement on what this

representation should be or how representation should be

derived. Several publications on the topic suggest that target

representation for such purposes be set between 20% and 40%

of the habitat’s historic distribution (Roberts and Hawkins,

1999; Ward et al., 1999; Turpie et al., 2000; Beck, 2003;

Groves, 2003).

This study aimed to identify a set of priority sites on which

to focus scarce management resources or conservation dollars.

While it was recognized that all areas are potentially important

for natural resource conservation, the goal in this study was to

identify areas to direct limited financial resources, staff and

time with the hope that conservation successes in these areas

could be used as leverage to achieve conservation in adjacent

areas or at a larger scale. As a means of identifying an

appropriate target representation approach, several

representation scenarios were explored, including two

universal representation scenarios (20% and 40%

representation for all conservation targets) and a variable

representation approach developed by The Nature

Conservancy (2003). In the variable target representation

approach, individual target representation was based on four

attributes: degree of rarity, vulnerability to human activities,

current status compared with historic and whether the target

represents a breeding site such as nesting colony or spawning

aggregation, etc. Three of the attributes (degree of rarity,

vulnerability to human activities, and current status compared

with historic) were rated on a scale of 1 to 3, with 1 being less

rare, vulnerable or compromised and 3 being more rare,

vulnerable or compromised. The breeding site attribute was

rated as either a ‘1’ for not a breeding site or a ‘3’ for breeding

site. All scores were based on information available in the

scientific literature. Representation for each target was then

assigned, ranging from 20% to 100%, based on the total

attribute score (Figure 2).

Boundary length

Boundary length or perimeter of the MARXAN solution

(boundary of all solution sites added together) can be modified

using the boundary length modifier (Wb), a constraint entered

into the MARXAN algorithm. Adjustment of this variable

affects the spatial arrangement and efficiency of the results.

MARXAN allows Wb to be set at values between 0 and 1. A

Wb of 0 returns the most spatially efficient result with no

consideration given to arrangement of the selected sites which

are typically small (single planning unit) and scattered

throughout the planning area. As the Wb value is increased,

overall site number is reduced and sites become more

aggregated (multiple planning unit sites). Adjustment of this

variable allows for the development of solutions that make

more sense from a practical conservation or resource

management perspective as attempting to manage thousands

of small sites scattered throughout the planning area would be

difficult and expensive. A Wb value of 0.05 was used in the

analyses presented here.

Planning unit cost

MARXAN requires that each planning unit be assigned a base

cost as the algorithm seeks to minimize cost in its optimum

solution. The cost function in the algorithm may also be used

to vary the relative value of each planning unit depending on

their attributes. The effect of varying planning unit cost is that

those with a lower cost have a greater likelihood of being

selected in the MARXAN output all other factors being equal.

Two approaches for assigning values to planning units were

Figure 2. Assignment of variable representation factors. Representation of each target in each subregion was scored based on four attributes: rarityvulnerability, current status as compared to historic condition, and whether the distribution coverage represents a breeding aggregation.

Representation factors ranging from 20% to 100% were assigned based on the frequency distribution of target scores.

IDENTIFICATION OF PRIORITY CONSERVATION SITES IN MARINE AREAS OF FLORIDA

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

examined. In the first approach, all planning units were valued

equally, and assigned a base cost of 85 points. In this type of

approach, where all planning units are valued equally, the

value of the planning unit cost selected is not important as long

as it is greater than 0 and is in balance with the species penalty

factor (discussed below).

The second approach to assigning costs to planning units,

involved the development of a spatial index of socio-economic

factors. The purpose of this approach was to better inform

marine resource managers and conservation practitioners

where priority attention could be focused while minimizing

conflicts with existing resource users and avoiding more

impacted locations. The spatial socio-economic index was

composed of activities not likely to be reversible or reversible

only at a high to very high cost. Only socio-economic uses that

have a demonstrated adverse impact on native habitats and

species (e.g. development, high intensity use and pollution) and

may be described as structures, facilities or activities were

considered for inclusion. Eleven such socio-economic uses

impacting marine and estuarine habitats in Florida where

geospatial information was available were identified. The 10

socio-economic factors selected and the scoring used to

describe the inlevel of impact are listed in Table 2 and

include population and road density, port facilities, major

shipping lanes, hardened shorelines, Superfund sites, major

National Pollutant Discharge Elimination System permitted

point source (NPDES) discharges, marine facilities and boat

ramps, offshore dredged disposal sites, and dredged shipping

channels.

Relative scoring of the socio-economic factors was

developed and refined through an iterative expert review

process that included marine resource managers, scientists and

conservationists working in Florida. Point locations of the

most concentrated adverse impacts (e.g. major municipal

discharges and Superfund sites) were assigned the highest score

of 500 points. More dispersed, but still intensive impacts (i.e.

shipping lanes with greater than 272 155 metric tons annual

uptonnage) were assigned 250 points along the entire shipping

lane length. The most dispersed socio-economic factors (e.g.

population, roads, hardened shoreline) were assigned scores

ranging from 10 to 100 points per unit of measure. Scores for

these most dispersed socio-economic factors are lower because

multiple factor units may occur within a single planning unit

(see Table 2 for the range of units for each socio-economic

factor). For example, of the planning units with the highest

socio-economic index scores (total score ranging from 4000 to

6212), none contain a Superfund site, major wastewater

discharge or offshore dredged disposal site. The high socio-

economic scores for these planning units are derived primarily

from population density, road density, and hardened

shoreline. A map depicting the socio-economic index is

shown in Figure 3. There are other threat factors that would

have been advantageous to include in this analysis if the time

and resources had been available, including trawling intensity,

observed seagrass propeller scarring and coastal watershed

impacts, etc. Where the described socio-economic uses

were present, the cost points outlined in the socio-economic

index were added to the base cost of 85 points for each

planning unit.

Species penalty factor

A species penalty factor is set for each conservation target and

represents the cost added to the total portfolio cost if the

conservation target goal is not met. Setting a high species

penalty helps to ensure that MARXAN will meet conservation

target goals. A species penalty factor of 500 was applied to all

conservation targets for all scenarios presented here.

Algorithm application

For each MARXAN run, the number of iterations was set at

10 million to ensure a high degree of confidence in optimizing

the conservation goals. Each run was repeated 100 times and

MARXAN selected the best of these configurations as the

optimal solution.

RESULTS

Two aspects of MARXAN output were evaluated, total

portfolio area selected and efficiency at meeting target goals,

Table 2. Socio-economic use index - factors and scoring for each planning unit

Socio-economic factors Unit Index Points Data source

Most Concentrated Impacts:Major NPDES� discharges Presence in planning unit 500 NOAA-OPISSuperfund sites (range, 1–2) For each site 500 NOAA-OPISOffshore dredge disposal sites (range, 1–2) For each site 500 NOAA-OPISMore Dispersed Impacts:Major shipping lanes (uptonnageX272,155metric tons)

Presence in planning unit 250 USACE - NavigationData Center (NDC)

Most Dispersed Impacts:Hardened shoreline (range, 0.003–115km) For every 2.5 kilometres 100 FWC/FWRIDredged shipping channels (range, 1–15) For each dredging project 25 USACE- NDCPort facilities (range, 1–31 facilities) For each facility 20 USACE- NDCPopulation density (range, 0.000218/km2

–18998/km2)Road density (range, 0.004–246km)

For each 38.6 people persquare kilometre

10 U.S. CensusCensus 2000

For each kilometre 10 U.S. CensusCensus 2000

Marine facilities/boat ramps (range: 1–33) For each facility/boat ramp 10 FWC/FWRI

Note: Socio-economic index points were added to the planning unit base cost of 85 to derive the total planning unit cost. �NPDES: NationalPollutant Discharge Elimination System (http://cfpub.epa.gov/npdes/).

L. GESELBRACHT ET AL.

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

to assess the influence of the algorithm parameters that were

varied (conservation targets, their representation and planning

unit cost). The efficiency of meeting target goals for each

scenario was calculated by averaging the percentage of target

goals met in each subregion, then averaged across all

subregions. Table 3 illustrates how this was done for

MARXAN scenario 2. The conservation targets listed in

Table 1 were included in all of the MARXAN runs presented

in this section with the exception of benthic complexity, which

was included in only one of the scenarios to illustrate its

influence on the MARXAN output. Likewise, the socio-

economic index was only included in one scenario, scenario 5,

to illustrate its influence on results. A preferred alternative was

selected from among the scenarios based on analysis of the

MARXAN results and expert review comments, before

conducting a spatial comparison of the preferred alternative

with existing managed areas.

Target goals

Under the universal target goal approach, the total amount of

priority area selected in the optimal solution increased as

target goals were increased. When conservation targets were

universally represented at 20% and 40%, 2.9% versus 5.6% of

the total planning area was selected (Table 4). The portfolio

resulting from using a 20% target goal is for the most part a

subset of the 40% goal portfolio. The only exception to this is

the southern tip of the Florida peninsula. The 20% target goal

scenario selected all of Biscayne Bay whereas the 40% goal

scenario selected northern and southern areas of Biscayne Bay

and a large area of Florida Bay. The variable target goal

approach delivered an intermediate result of 5.0% of the total

planning area selected. Differences between areas selected in all

three scenarios are modest (Figure 4(a)–(c)). When benthic

complexity is not included as a conservation target, most

target distribution included in this study are nearshore and

coastal. If the planning area was limited to state waters (3

nautical miles off the Atlantic Coast and 9 nm off the Gulf of

Mexico coastline) where target data density is high, the total

amount of planning area selected for the 20%, 40% and

variable target representation scenarios is as follows: 18.6%,

36.3% and 31.6%, respectively.

Regarding efficiency of meeting the target goals, goals were

achieved for all of the conservation targets in all of the

scenarios presented in this study. Efficiency of meeting target

goals, however, differed among the scenarios. Thirty percent

over target goals was selected as an arbitrary limit for

acceptable representation and anything over this 130% value

was considered as excessive over-representation. Under the

variable target goal approach, 51% of the targets exceeded

their goals by 30% or more. Under the universal goal

approaches of 20% and 40%, the portion of targets

exceeding 30% of their goals was 62% and 46%, respectively

(Table 4). Therefore, the 40% universal target goal approach

Figure 3. Spatial Index of socio-economic factors. The spatial index is a relative scoring of the following socio-economic factors not likely to bereversible or reversible only at a high to very high cost: population and road density, port facilities, major shipping lanes, hardened shorelines,Superfund sites, major permitted point source discharges, marine facilities and boat ramps, offshore dredged material disposal sites, and dredgedshipping channels. The darkest areas represent the highest concentration of these socio-economic factors. Table 2 provides the scoring methodology.

IDENTIFICATION OF PRIORITY CONSERVATION SITES IN MARINE AREAS OF FLORIDA

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

delivered the most efficient results in terms of meeting target

goals among the three approaches examined. The variable

target goal approach was viewed less favourably by the expert

review panel owing to the necessity to employ some

subjectivity in rating criteria for some conservation targets.

For several targets, quantitative information on pre-settlement

extent or population size is limited or unavailable (Geselbracht

et al., 2005).

Influence of including benthic complexity

The effect of including the surrogate target, benthic complexity

was examined by comparing scenarios that were identical

except for their inclusion/omission of benthic complexity,

using the scenario with a 40% universal target goal. When

benthic complexity was included, results for the offshore

portions of the planning area (here defined as greater than

Table 3. Calculation of efficiency at meeting target goals for marxan scenario 2 (40% universal target goal; benthic complexity excluded)

Percentage target goals captured by subregion:

Conservation targets 1 2 3 4 5 6 7 8

Patch coral reef� na�� na na 103% na na na naShallow bank coral reef� na 250% 114% 117% na na na naDeep bank coral reef� na 250% 103% 100% na na na naDeep reef resources� 239% 102% 103% 103% na na na naMangrove swamp 100% 102% 117% 128% 100% 121% 115% naBeach/Surf zone 104% 102% 123% 203% 138% 100% 133% 103%Salt marsh 148% 127% 142% 174% 131% 126% 105% 106%Submerged aquatic vegetation 128% 113% 216% 100% 120% 101% 100% 145%Coastal tidal river or stream 100% 114% 108% 119% 165% 112% 115% 103%Tide flat 157% 116% 245% 101% 103% 100% 101% 101%Bivalve (oyster) reef 159% 250% na na 208% 173% 134% naAnnelid (worm) reef na 225% 155% na na na na naHardbottom na 112% 122% 131% na na na naInlets and passes 190% 160% 116% 125% na 101% 104% 107%Florida manatee 202% 213% 137% 120% 141% 188% 250% naRight whale calving grounds 100% 113% na na na na na naAmerican oystercatcher 159% 100% na na 208% 100% 116% 250%Black skimmer 208% 179% 250% 250% 200% 104% 200% 133%Brown pelican 103% 138% na 105% 125% 110% 125% 250%Least tern 111% 156% 125% 125% 208% 125% 208% 117%Piping plover 100% na 250% 125% 125% 103% 125% 150%Reddish egret 125% 188% 250% 160% na 117% 250% 250%Roseate spoonbill 150% 167% 250% 201% na 107% na naRoseate tern na na na 100% na 250% na naSnowy egret 125% 115% 167% 155% 125% 128% 167% 150%Snowy plover na na na 250% 250% 113% 114% 145%Wading bird colony 125% 167% 250% 104% 167% 104% 167% naAmerican crocodile na na 129% 154% 262% 120% na naGreen turtle nesting beaches 188% 113% 131% 135% na 239% 166% 100%Loggerhead turtle nesting beaches 166% 103% 177% 100% 113% 102% 100% 108%Leatherback turtle nesting beaches 101% 158% 214% na na na 154% naHawksbill turtle nesting beaches na na na 250% na na na naKemp’s Ridley nesting beaches 129% na na na na na na naTurtles in-water na 106% 129% 100% na na na naOrnate diamondback terrapin na na na 100% 108% 119% 122% 123%Mississippi diamondback terrapin na na na na na na na 116%Carolina diamondback terrapin 242% na na na na na na naSmalltooth sawfish 125% 125% 250% 109% 118% 103% 111% naSlashsheek goby na 200% 250% na na na na naRiver goby na 188% na na na na na 250%Bigmouth sleeper na 150% 250% na na na na naMangrove Rivulus na 109% 125% 167% na 167% na naOpossum pipefish 125% 132% 250% na na 250% na naStriped croaker na 250% na na na na na naGulf sturgeon na na na na na 125% 250% 125%Atlantic sturgeon 250% 250% na na na na na naSaltmarsh topminnow na na na na na na na 107%Alabama shad na na na na na na 161% 231%Key silverside na na na 120% na na na naAcropora palmata na na 167% 237% na na na naJohnson’s seagrass na 110% 126% na na na na naSpawning aggregations na na 250% 103% na na na naPercentage targets with goals 4130% in subregion: 46% 54% 63% 38% 55% 21% 48% 45%Percentage targets with goals 4130% statewide: 46%

�The coral reef conservation target was split into the reef types listed above. ��na5not applicable. Target was not present in subregion.

L. GESELBRACHT ET AL.

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

50m) differed substantially, as anticipated, because the benthic

complexity dataset was the only dataset used in this study that

covered the entire offshore portions of the planning area

(Figures 4(b) and (d)). Many of the offshore areas gained, as a

result of including benthic complexity, coincide with known

deep reef features of high biodiversity value including Pulley

Ridge, Pourtales Terrace, Miami Terrace, Steamboat Lumps and

DeSoto Canyon (Koenig et al., 2000; Scanlon, 2000; Scanlon et

al., 2001; Sheridan and Caldwell, 2002; Gardner et al., 2003;

Jarrett et al., 2004; Reed, 2004). Differences in results between the

nearshore and coastal portions of these two scenarios were

modest with four exceptions. The scenario excluding benthic

complexity (4b) includes a large site in Charlotte Harbor. The

scenario with benthic complexity (4d) includes two large sites

along the Florida Panhandle and one large site in Tampa Bay. In

addition, the scenario with benthic complexity favoured Biscayne

Bay over Florida Bay and split the large Panhandle sites into

several small sites. Closer examination of why including benthic

complexity may have influenced some of the results in this way

revealed that in the shallower portions of the planning area,

benthic complexity highlighted man-made features such as

shipping lanes and other dredged areas.

The scenario that included benthic complexity (4d) selected

more than twice the area of the scenario without benthic

complexity, reflecting greater spatial representation in the

offshore area. In terms of efficiency of meeting conservation

target goals, all goals were met with 54% of targets exceeding the

target goal by 30% or more compared with 46% for the scenario

without benthic complexity (Table 4). Therefore, the scenario

with benthic complexity was less efficient at meeting target goals

than the identical scenario without benthic complexity.

Effect of including socio-economic use index as a cost

surface

The effect of including a spatial index of socio-economic activities

or threats was explored to examine its influence on selection of

priority areas. A comparison of the optimal portfolio of sites,

both with and without the socio-economic index included, shows

many similarities (Figures 4(b) and (e)). A large site off extreme

north-east Florida was selected by both approaches as were sites

in the Mosquito, Banana and Indian River lagoons (subregions 1

and 2). The scenario including the socio-economic index (4e)

differs from the scenario excluding it (4b) in the following ways:

the Florida Panhandle sites are more fragmented, there is a gap in

the selected areas at Crystal River in the southern Big Bend

Subregion, and smaller sites in Florida Bay and the St Lucie

Estuary were selected. In the portfolio with the socio-economic

index included, results appear to avoid more developed or higher

cost areas.

In terms of efficiency, the scenario including the socio-

economic index exhibits improved efficiency both spatially and

at meeting target goals. This scenario selected 4.8% of the total

planning area, all conservation target goals were met and 38%

(versus 46%) of targets exceeded the 40% target goal by 30%

or more (Table 4). While the expert review panel recognized

the value of a socio-economic index for informing

conservation and resource management activities, they

expressed a distinct preference for selecting priority sites

based solely on biodiversity importance rather than giving

preference to sites less subject to human use/disturbance. Their

rational was that important biodiversity sites should not be de-

emphasized if they were also important for socio-economic

uses. Rather the expert review panel suggested that sites

important for both biodiversity values and socio-economic

uses may warrant more concentrated management attention

rather than less. They recommended using the socio-economic

index to guide resource management and conservation

activities following identification of priority sites.

Identification of a preferred portfolio and comparison with

existing managed areas

Based on the analysis of the MARXAN output and the general

advice of our expert review panel, a preferred portfolio was

selected that represented all targets at the 40% level, included

benthic complexity as a target only in deeper reaches of the

planning area (450m depth) and did not include the socio-

economic index in site selection. The results of this scenario are

presented in Figure 4(f). The preferred scenario selected 9.7%

of the planning area, more than the scenarios that completely

excluded benthic complexity, but less than the scenario that

included it throughout the planning area. Efficiency at meeting

target goals was intermediate (48%) compared with the other

scenarios presented in this study.

A spatial comparison of the preferred alternative with

existing managed areas was undertaken to evaluate the extent

to which preferred alternative sites fall within some type of

management regime (Figure 5). Approximately 39% of the

preferred alternative falls into some form of existing managed

area. The overlapping managed areas range from seasonal

fishing closure areas to no take marine reserves.

DISCUSSION

The site selection algorithm, MARXAN, greatly simplifies

identification of a spatially efficient portfolio of priority areas

Table 4. Summary of MARXAN output statistics

MARXAN scenario Percentage ofplanning areaselected

Percentageof targets over-represented by430% of goal

1. 20% universal target goal,benthic complexityexcluded, BLM5 0.05

2.9% 62%

2. 40% universal target goal,benthic complexityexcluded, BLM5 0.05

5.6% 46%

3. Variable target goal,benthic complexityexcluded, BLM5 0.05

5.0% 51%

4. 40% universal target goal,benthic complexityincluded, BLM5 0.05

13.4% 54%

5. 40% universal target goal,benthic complexityexcluded, BLM5 0.05,socio-economic indexincluded.

4.8% 38%

Preferred alternative.40% universal target goal,benthic complexity only inareas 450m, BLM5 0.05;socio-economic indexexcluded

9.7% 48%

IDENTIFICATION OF PRIORITY CONSERVATION SITES IN MARINE AREAS OF FLORIDA

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

using spatial information on conservation targets and objective

criteria on desired levels of representation. It is an excellent

tool that informs a complex process and enables those

interested in identifying priority sites to observe how changes

in selection criteria influence portfolio design. A good

knowledge of how input parameters influence portfolio

design will greatly enhance the ability of MARXAN users to

satisfactorily integrate the expert advice of scientists, resource

managers and the sometimes competing concerns of other

stakeholders. Identification of priority sites using only expert

input has been described as less defensible and scientifically

credible (Lourie and Vincent, 2004). However, development of

priority sites using biogeographic data and a site selection

algorithm without stakeholder input may be equally

problematic. Employing quantitative information and

decision-support tools such as MARXAN, while soliciting

expert and other stakeholder input throughout the process, is

likely the best approach to priority site identification. Broader

support of the process and results are built concurrently with

development of the priority site portfolio.

Setting target goals can be a difficult exercise due to the

varying opinions on how it should be done and the lack of well

established guidance. Employing universal target goals is an

expedient method. Scenario results can be compared on the

basis of both spatial efficiency and efficiency at meeting the

target goals. Although the effort to set goals individually for

each conservation target was not popular among the expert

reviewers, refining this approach has merit. Setting all target

goals at the same level does not account for differences in

historical exploitation levels or the shifting of baseline

conditions. Refining an approach that sets representation

goals individually for each target will require advancing our

understanding of historical distribution of conservation

targets. Pauly (1995) supports improving our understanding

of historical distributions as a means for setting more

appropriate conservation target goals.

Inclusion of surrogate conservation targets in MARXAN

analyses may return some unexpected results. In shallower

portions of the planning area, benthic complexity highlighted

areas where shipping channels and dredged sites were

present. These high impact sites are not typically considered

of high biodiversity value, so the benthic complexity target was

excluded from areas with depths less than 50m in the

preferred alternative. This result in shallow portions of

Florida marine and estuarine areas may also be a function of

Florida’s predominantly low relief environment with the

exception of coral reef areas. Benthic complexity does,

however, appear to be a valuable surrogate target to

include in the deeper portions of the planning area and

may aid in the identification of areas of potential conservation

interest that have not been subject to extensive

scientific investigations. Use of benthic complexity as a

target in site selection may best be confined to deeper

portions of marine planning areas where conservation target

Figure 4. Results of site prioritization analyses using MARXAN with several different scenarios. Expert review of the various scenario outputsresulted in the creation of the preferred scenario, which can be used to guide future investment in conservation and management activates. All runs

used a boundary length modifier of 0.05 and a species penalty factor of 500.

L. GESELBRACHT ET AL.

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

distribution data are limited, or to areas where man-made

features are not likely to influence the results.

The purpose of conducting the site prioritization effort will

drive the extent to which socio-economic information is

incorporated. The emphasis of this effort was to identify a

portfolio of sites representing high biodiversity value.

Participants in the expert review sessions were heavily

weighted towards marine scientists, government resource

managers and members of conservation organizations. There

was little representation from user groups. Representation by

user groups under these conditions was not essential as the

process did not result in any actual proposals for new marine

reserve sites, which would have resulted in limiting the access

of some user groups. The use of socio-economic data to

identify sites of high use was much less an issue here than in

other similar reserve design efforts. Subsequent efforts to

identify and implement appropriate conservation strategies for

priority sites will require the use of socio-economic

information and the involvement of user groups, especially if

any restrictions on site use are proposed.

While it was found that approximately 39% of the

preferred alternative falls within the boundaries of existing

managed areas, this should not be construed as 39% successful

conservation of the preferred portfolio. The existing managed

areas overlapping the preferred alternative sites vary greatly in

regulatory purview, staffing, enforcement, etc. Only the no-

take marine reserves, which represent less than 1% of preferred

alternative area, provide protection from most of the direct

human use impacts. The majority of the managed areas

overlapping the preferred portfolio provide incomplete

protection, which may include seasonal or permanent

prohibitions on some but not all types of fishing and/or

limited protection from habitat destruction. Establishing an

appropriate managed areas framework from which to launch

effective, long-term conservation strategies for the marine

systems in Florida’s and surrounding waters will likely require

expanding some existing managed areas, creating new

managed areas and strengthening regulatory protection so

that at the very least direct site impacts are minimized.

Identifying priority areas will always be controversial, but

avoiding this kind of analysis risks losing the systems we seek

to protect (Lourie and Vincent, 2004). Marine conservation in

Florida has reached a crossroads. While there is an extensive

system of state, federal, local and private marine managed

areas in state and adjacent marine waters, many of these sites

operate independently of one another and most offer only

limited protection to marine resources. The goals and

strategies of these managed area programmes should be

aligned, quantitatively articulated and their regulatory

authorities strengthened to ensure the long-term health of

marine systems in and around Florida’s waters for the use,

enjoyment and benefit of future generations. More generally,

quantitative explorations of the effectiveness of tools like

MARXAN in specific case studies have represented a gap in

Figure 5. Preferred alternative is compared with existing managed areas in the study area. Dark shading indicates permanently designated managedareas. Light shading indicates areas only subject to temporary or seasonal fishing closures.

IDENTIFICATION OF PRIORITY CONSERVATION SITES IN MARINE AREAS OF FLORIDA

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

our current body of knowledge. The methodologies and results

presented in this paper can be used by resource managers,

conservation professionals and other stakeholders to help

define how marine resource management can be strengthened

and to identify where scarce conservation and management

funds may best be focused in a particular planning area.

ACKNOWLEDGEMENTS

Funding for this project was provided by the US Department

of Interior, State Wildlife Grants Program T-4 Grant

administered by the State of Florida and by The Nature

Conservancy. We are grateful for the contributions of the

many participants and reviewers of this effort which are too

numerous to list here. A portion of the funding for this project

was provided by the US Fish and Wildlife Service State

Wildlife Grants Program T-4 Grant administered by the

Florida Fish and Wildlife Conservation Commission (FWC

Contract No. 04122).

REFERENCES

Ardron J. 2002. A recipe for determining benthic complexity:an indicator of species richness. In Marine Geography: GISfor the Oceans and Seas, Breman J (ed.). ESRI Press:Redlands, CA; 169–175.

Ball I, Possingham H. 2000. MARXAN (V1.8.2): MarineReserve Design Using Spatially Explicit Annealing, a Manual.The Ecology Centre, University of Queensland: Brisbane.

Beck M. 2003. The sea around: marine regional planning. InDrafting a Conservation Blueprint: a Practitioners’ Guide toPlanning for Biodiversity, Groves CR (ed.). Island Press:Washington, DC; 319–344.

Beck M, Odaya M. 2001. Ecoregional planning marineenvironments: identifying priority sites for conservation inthe northern Gulf of Mexico. Aquatic Conservation: Marineand Freshwater Ecosystems 11: 235–242.

Davis Jr R. 1997. Geology of the Florida Coast. In TheGeology of Florida, Randazzo A, Jones D (eds). Universityof Florida Press: Gainesville; 155–168.

DeBlieu J, Beck M, Dorfman D, Ertel P. 2005. Conservation inthe Carolinian Ecoregion: An Ecoregional Assessment. TheNature Conservancy: Arlington, VA.

Floberg J, Goering M, Wilhere G, MacDonald C, Chappell C,Rumsey C, Ferdana Z, Holt A, Skidmore P, Horsman T, etal. 2004. Willamette Valley-Puget Trough-Georgia BasinEcoregional Assessment, Volume One: Report. Prepared byThe Nature Conservancy with support from the NatureConservancy of Canada, Washington Department of Fishand Wildlife, Washington Department of Natural Resources(Natural Heritage and Nearshore Habitat programs),Oregon State Natural Heritage Information Center and theBritish Columbia Conservation Data Centre.

Gardner JV, Hughes Clarke J, Mayer L. 2003. Bathymetry andAcoustic Backscatter of the Mid and Outer ContinentalShelf, Head of De Soto Canyon, Northeastern Gulf ofMexico—Data, Images, and GIS. US Geological SurveyOpen-File Report OF 02–396.

Geselbracht L, Torres R, Cumming G, Dorfman D, Beck M.2005. Marine/Estuarine Site Assessment for Florida: AFramework for Site Prioritization. The NatureConservancy: Gainesville, FL.

Groves C. 2003. Drafting a Conservation Blueprint: aPractitioner’s Guide to Planning for Biodiversity. IslandPress: Washington, DC.

Hockey PAR, Branch GM. 1997. Criteria, objectives andmethodology for evaluating marine protected areas inSouth Africa. South African Journal of Marine Science 18:369–383.

Jarrett BD, Hine AC, Halley RB, Naar DF, Locer SD,Neumann AC, Twichell D, Donahue CHBT, Jaap WC,Palandeo D, Cembronowicz . 2004. Strange bedfellows—adeep-water hermatypic coral reef superimposed on adrowned barrier island; southern Pulley Ridge SW Floridaplatform margin. Marine Geology 24: 295–307.

Koenig CC, Coleman FC, Grimes CB, Fitzhugh GR, ScanlonKM, Gledhill CT, Grace M. 2000. Protection of fishspawning habitat for the conservation of warm temperatereef fish fisheries of shelf-edge reefs of Florida. Bulletin ofMarine Science 66: 593–616.

Leslie H. 2005. A synthesis of marine conservation planningapproaches. Conservation Biology 19: 1701–1713.

Leslie H, Ruckelshaus M, Ball I, Andelman S, Possingham H.2003. Using siting algorithms in the design of marine reservenetworks. Ecological Applications 13: S185–S198.

Lodge T.2005. The Everglades Handbook–Understanding theEcosystem. CRC Press: Washington, DC.

Lourie S, Vincent A. 2004. Using biogeography to help setpriorities in marine conservation. Conservation Biology 18:1004–1020.

Madley KA, Sargent B, Sargent FJ. 2002. Development of aSystem for Classification of Habitats in Estuarine andMarine Environments (SCHEME) for Florida. Report tothe US Environmental Protection Agency, Gulf of MexicoProgram (Grant Assistance Agreement MX-97408100).Florida Marine Research Institute, Florida Fish andWildlife Conservation Commission, St. Petersburg.

Pauly D. 1995. Anecdotes and the shifting baseline syndromeof fisheries. Trends in Ecology and Evolution 10: 430.

Possingham HP, Ball IR, Andelman S. 2000. Mathematicalmethods for identifying representative reserve networks. InQuantitative Methods for Conservation Biology, Ferson S,Burgman M (eds). Springer-Verlag: New York; 291–305.

Pressey RL, Johnson IR, Wilson PD. 1994. Shades ofirreplaceability: towards a measure of the contribution ofsites to a reservation goal. Biodiversity and Conservation 3:242–262.

Randazzo A, Halley R. 1997. Geology of the Florida Keys. InThe Geology of Florida, Randazzo A, Jones D (eds).University of Florida Press: Gainesville; 251–260.

Reed JK. 2004. General description of deep-water coral reefsof Florida, Georgia and South Carolina: A summary ofcurrent knowledge of the distribution, habitat, andassociated fauna. Report to the South Atlantic FisheryManagement Council, NOAA, NMFS.

Roberts CM, Hawkins JP. 1999. Extinction risk in the sea.Trends in Ecology and Evolution. 15: 241–246.

Roberts CM, Branch G, Bustamante RH, Castilla JC, DuganJ, Halpern BS, Lafferty KD, Leslie H, Lubcenco J, McArdleD, et al. 2003a. Application of ecological criteria in selectingmarine reserves and developing reserve networks. EcologicalApplications 13: S215–S228.

Roberts CM, Andelman S, Branch G, Bustamante RH,Castilla JC, Dugan J, Halpern BS, Lafferty KD, Leslie H,Lubcenco J, McArdle D, et al. 2003b. Ecological criteria forevaluating candidate sites for marine reserves. EcologicalApplications 13: S199–S214.

Salm RV, Clark J, Siirila E. 2000. Marine and CoastalProtected Areas. A Guide for Planners and Managers, 3rdedn. IUCN: Washington, DC.

L. GESELBRACHT ET AL.

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc

Scanlon KM. 2000. Surficial seafloor geology of a shelf-edgearea off West Florida. In West Florida Shelf: Sidescan-sonarand Sediment Data from Shelf-edge Habitats in theNorthEastern Gulf of Mexico, Briere PR, Scanlon KM,Fitzhugh G, Gledhill CT, Koenig CC (eds) US. GeologicalSurvey, Open-file Report 99–589.

Scanlon KM, Koenig CC, Coleman FC, Rozycki JE. 2001.Paleoshorelines, drowned reefs, and grouper habitat in thenortheastern Gulf of Mexico. Geology of Marine HabitatSession, Geological Association of Canada Annual Meeting,2001, St. Johns, Vol. 26.

Shaffer ML, Stein BL. 2000. Safeguarding our preciousheritage. In Precious Heritage: The Status of Biodiversity inthe United States, Stein BA, Kutner LS, Adams JS (eds).Oxford University Press: Oxford; 301–322.

Sheridan P, Caldwell P. 2002. Compilation of datasets relevant to the identification of essential fishhabitat on the Gulf of Mexico continental shelf and for

the estimation of the effects of shrimp trawlinggear on habitat. NOAA Technical Memorandum NMFS-SEFSC-483

The Nature Conservancy, Greater Caribbean EcoregionalPlan Team. 2003. An Ecoregional Plan for Puerto Rico:Portfolio Design. Report to Bristol-Myers SquibbCompany.

Turpie JK, Beckley LE, Katua SM. 2000. Biogeographyand the selection of priority areas for conservationof South African coastal fishes. Biological Conservation 92:59–72.

US Fish and Wildlife Service. 1979. Classification of wetlandsand deepwater habitats of the United States. BiologicalServices Program, FWS/OBS-79-31.

Ward TJ, Vanderklift MA, Nicholls AO, Kenchington RA.1999. Selecting marine reserves using habitats and speciesassemblages as surrogates for biological diversity. EcologicalApplications 9: 691–698.

IDENTIFICATION OF PRIORITY CONSERVATION SITES IN MARINE AREAS OF FLORIDA

Copyright r 2008 John Wiley & Sons, Ltd. Aquatic Conserv: Mar. Freshw. Ecosyst. (2008)

DOI: 10.1002/aqc