assessing state-wide biodiversity in the florida gap ... · assessing state-wide biodiversity in...

Journal of Environmental Management (2002) 66, 127±144doi:10.1006/jema.2002.0551, available online at http://www.idealibrary.com on

1

Assessing state-wide biodiversity in theFlorida Gap analysis project

L. G. Pearlstine ²*, S. E. Smith ³, L. A. Brandt §, C. R. Allen¶,W. M. Kitchens k and J. Stenberg#

² Assistant Scientist, Department of Wildlife Ecology and Conservation,University of Florida, Gainesville, FL 32611, USA³ Associate Professor, Department of Civil Engineering, University of Florida, Gainesville,FL 32611, USA

§ Senior Wildlife Biologist, A. R. M. Loxahatchee National Wildlife Refuge,U.S. Fish and Wildlife Service, Boynton Beach, FL 33437, USA¶ Unit Leader, South Carolina Cooperative Fish and Wildlife Research Unit,Clemson, SC 29634, USAk Research Scientist, Florida Cooperative Fish and Wildlife Research Unit,U.S. Geological Survey, Gainesville, FL 32611, USA# Environmental Scientist III, St Johns River Water Management District,Palatka, FL 32178, USA

Received 27 April 2001; accepted 2 January 2002

The Florida Gap (Fl-Gap) project provides an assessment of the degree to which native animal species and naturalcommunities are or are not represented in existing conservation lands. Those species and communities not adequatelyrepresented in areas being managed for native species constitute `gaps' in the existing network of conservation lands.The United States Geological Survey Gap Analysis Program is a national effort and so, eventually, all 50 states will havecompleted it. The objective of Fl-Gap was to provide broad geographic information on the status of terrestrial vertebrates,butter¯ies, skippers and ants and their respective habitats to address the loss of biological diversity. To model thedistributions and potential habitat of all terrestrial species of mammals, breeding birds, reptiles, amphibians, butter¯ies,skippers and ants in Florida, natural land cover was mapped to the level of dominant or co-dominant plant species. Landcover was classi®ed from Landsat Thematic Mapper (TM) satellite imagery and auxiliary data such as the nationalwetlands inventory (NWI), soils maps, aerial imagery, existing land use/land cover maps, and on-the-ground surveys.Wildlife distribution models were produced by identifying suitable habitat for each species within that species' range.Mammalian models also assessed a minimum critical area required for sustainability of the species' population. Wildlifespecies richness was summarized against land stewardship ranked by an area's mandates for conservation protection.

# 2002 Elsevier Science Ltd. All rights reserved.

Keywords: biodiversity, land cover classi®cation, habitat modeling, gap analysis program, Florida.

Introduction

Florida is a state with diverse and unique speciesand landscapes. Its geographic position, spanning

0301±4797/02/$ ± see front matter

* Corresponding author. E-mail: pearlstn@u¯.edu

latitudes from temperate to semi-tropic climates,plays a major role in shaping its biotic assemblages.Florida is mainly a large peninsula of NorthAmerica that extends about 650 km between theAtlantic Ocean and the Gulf of Mexico. Theserelatively warm climatic in¯uences create humidconditions that support lush and diverse vegetation.

# 2002 Elsevier Science Ltd. All rights reserved.

128 L. G. Pearlstine et al.

Even the central portions of the peninsula andpanhandle are in¯uenced by tropical conditionsthrough southerly wind patterns.

Anthropogenic disturbance from populationgrowth is threatening the existence of some Floridaecosystems and the sustainability of many others.Between 1960 and 2000, the population of the statemore than tripled. In the decade between 1990 and2000 the population of Florida increased by morethan 3 million. (Bureau of Census, 1960; Bureau ofCensus, 1990; Bureau of Census, 2000). The Federalgovernment owns only about 9% of Florida's land-mass with state and other government entitiesowning very small amounts of land. The rest isprivately held.

Expanding human habitation comes at theexpense of natural habitat. It has resulted in habitatloss, fragmentation, con¯ict over water needs anduse, and increasing recreational pressures on theecosystems. When the Florida Fish and WildlifeConservation Commission published the ®rst state-wide analysis of land covers, wildlife distributionsand areas needing conservation protection in 1994(Cox et al., 1994), less than 10% of the State's histo-rical longleaf pine natural community remained,and less than 5% of the State's rare pine rocklandsnatural community remained. Florida is still diversein both plants and animals, but the persistence ofhabitats will depend on better and scienti®callybased planning for both resources and growthtogether with conservation action.

The mission of the US National Gap AnalysisProgram (Gap) is to mitigate wildlife conservationproblems by `providing an assessment of the essen-tial biotic elements (plant communities and nativeanimal species) and to facilitate the application ofthis information to land management activities'(Scott et al., 1987; Scott and Jennings, 1994). Gapanalysis relies on maps of dominant natural landcover types as the most fundamental spatial com-ponent of the analysis (Scott et al., 1993) forterrestrial environments.

Methods

Land cover classi®cation

Land cover was classi®ed to an aggregation of theNational Vegetation Classi®cation Scheme (NVCS)(The Nature Conservancy, 1997). The NVCS is anecologically based, hierarchical classi®cation thattreats all existing terrestrial vegetation types in onesystem. The Nature Conservancy (TNC) and the

Natural Heritage Network (Grossman et al., 1994)have been improving upon this system. The basicassumptions and de®nitions for this classi®cationsystem have been described by Jennings (1993). Thebasis for aggregating of the NVCS for Gap analysisin the Southeast United States is presented inPearlstine and McKerrow (1999).

Classi®cation was accomplished using 1992±94Landsat Thematic Mapper (TM) satellite imageryand auxiliary sources of data as described below andin Figure 1. The classi®cation approach involvedpreparation and transformation of the imagery,strati®cation, and iterative, unsupervised classi®ca-tion. Most of the scenes used to cover the state ofFlorida are from the winter of 1993 or spring 1994.Images for two of the scene areas were availablefrom both spring and winter, and so were used in amulti-temporal analysis to aid the classi®cation.

Water Management District land use/land covermaps were overlaid onto the satellite images tocheck the geometric consistency of the images. Ifpositional errors were present, an af®ne transform-ation and nearest-neighbor resampling were used toco-register the imagery with the land use/land covermaps.

Atmospheric haze existed in borne of the imagesand was removed in the pre-processing stage ofthe image processing. Crist et al. (1984) presenteda technique that we used for minimizing the effectsof haze by subtracting the image from the fourthspectral band of a tasseled cap transformation. Thetasseled cap transformation's fourth band corres-ponds with haze features present in the image andhas little or no effect on portions of the image wherehaze is not present.

The ®rst three spectral bands of the tasseled captransformation correspond to `brightness', `green-ness' and `wetness' in the image and have beenshown to be effective for improving classi®cationresults (Crist, 1984). These three transformedbands and bands 2, 3, 4, and 5 of the original TMimage were combined for the ®nal image used forclassi®cation.

For the two scenes with spring and winter images,the images were co-registered and normalizedbefore being combined. Normalization was neces-sary to correct for differences in sensor offset andgain, and scene illumination. The difference inoverall brightness between the images was normal-ized using a linear image regression process as des-cribed in the ERDAS Field Guide (ERDAS, 1999)and Jensen (1996).

This approach is well suited for multi-temporalanalysis where care must be taken not to adjustthe image for the seasonal variation of vegetation.

Check Registration AgainstMasking Layer TM

Image 1

Check & correct image(s) foratmospheric haze as needed

Co-register images,Normalize (multi-date only)

Compute Tasseled CapBands 1-3 for image(s)

Combine TM1 Tasseled Capbands (1,2,3) with TM2Tasseled Cap bands (1,2,3)

(multi-date only)

TMImage 2

MaskingLayer

TasseledImage

Classified Natural Areas(8 classes)

(masked and embedded)

Classified Urban Areas(10 to 30 classes)

Using urban mask codes

Classified Ag Areas(10 to 30 classes)

Using agriculture mask codes

Visually interpret classified urban & ag imagesSelect embedded natural areas

Class1

Class2

Class3

Class4

Class5

Class6

Class7

Class8

Under Each Class, Create 5 to 10 New Classes

Land Cover Mapping

Stratify classified naturalareas image

StratificationLayer

Stratified NaturalAreas Image

Put Final Labels onClasses –

Land Cover Map

UnsupervisedClassification ofNatural Areas50 to 80 Classes

Summarize ClassesAgainst GroundInformation

Combine/Bust Classesas Needed

Iterate

Figure 1. Land cover mapping ¯ow chart.

Assessing biodiversity in Florida 129

Fresh

Salt

PatchA

PatchB

PatchC

Figure 2. Retention of community shape when stratifyinga classi®ed satellite image by another dataset. In thissynthetic ®gure, National Wetlands Inventory (NWI)mapped data is used to stratify areas as either freshwaterof saltwater communities. The land cover type in Patch A isentirely within the NWI freshwater boundary. Patch B,however, falls across the NWI boundary. Because themajority of the patch is within saltwater, the whole patch isconsidered to be a saltwater community. Likewise, Patch Cis entirely de®ned as freshwater even though part of thepatch extends beyond the NWI boundary. In this fashion,the ®nal boundary between fresh and saltwater commu-nities generally follows the drawn NWI information, but isin¯uenced by the found patterns classi®ed from thesatellite image.


A regression model to account for these differenceswas created by ®rst identifying several `bright' and`dark' objects in each scene and, for each band,recording the digital number (DN). The darkestpixel was assigned the DN `zero'. Examples of `dark'objects were uniform non-turbid man-made lakesand coniferous forests. Examples of `bright' objectswere airport runways, large roads, beaches, and dryexposed soils.

Once these values were compiled, a linear regres-sion model was computed with the darker of the twoimages (i.e., the image with the overall lower aver-age DN for all bands and all pixels) was assigned tothe x variable. No negative numbers at the pixel-to-pixel level existed due to the fact that the darkestpixel's DN was set to zero. This insured that positivecorrections were made such that when applied, nonegative numbers resulted in the output image. Foreach band, a linear regression model and an associ-ated scatter plot were computed. If the model hada correlation coef®cient (r) higher than 95% and thescatter plot did not have signi®cant outliers, thelinear model was used. When outliers were detected,they were removed and the regression model wasrecomputed.

Urban and agricultural areas of a satellite imagetypically have a much wider variance of DN valuesthatarefoundinnaturalareas.Asaresult,signaturesdescribing urban and agricultural areas can oftenobscure or confuse discrimination of natural vegeta-tiontypes.Tominimizethoseeffects, theimageswerestrati®ed in `natural areas' and `developed areas' bymasking out urban and agricultural areas with stateWater Management District land use/land covermaps. Becuase these land use maps often delineate`developed' classes that contain embedded naturalareas that we wanted to retain, the developed areaimage was classi®ed using the `ISODATA' routine inthe ERDAS remote sensing software (ERDAS, 1999).Classes that appeared to represent natural areasoccurring within the developed areas were identi®edand the spectral image under those classes wasreassigned to the natural areas image.

The natural areas image was strati®ed once morebefore beginning classi®cation. This strati®cationtook advantage of existing soils and wetlands mapsto further reduce the variance of image DN valuesthat are considered together. USDA National SoilSurvey Center digital county-level soil maps wereaggregated to 13 broad classes. US Fish and WildlifeService National Wetlands Inventory maps wereaggregated to 18 classes. Neither map series coveredthe entire state, so, depending on availability, one orthe other or both were used for strati®cation over anentire scene. Using these digital vector coverages

directly to mask out parts of the image would createa classi®cation with hard boundaries that are arti-facts of the vector coverage rather than changes inthe re¯ectance values in the image. In order tocreate more realistic boundaries between classes,contiguous, spectrally similar pixels were treated asa group (Figure. 2). To accomplish this, an unsuper-vised classi®cation was performed on the entirenatural areas image, creating a classi®ed image with50 to 80 classes. When the map used for strati®ca-tion was laid over the classi®ed image, contiguouspixels classi®ed to the same value that fell with amajority in one strata stayed together with thatstrata rather than being split on the boundary line.This then became the mask to split the image intomany separate natural area images, each represent-ing the part of a single scene under one strata.

The ®nal step was the actual classi®cation of theimage subscenes into labeled vegetation classes.This was an iterative process of `ISODATA' cluster-ing and minimum distance classi®cation. Followingclassi®cation, labeling of the resulting classes was®rst attempted with the assistance of ground truthinformation, auxiliary data sources, low altitudeaerial videography and aerial digital imagery. Whena class represented more than one predominate


vegetation type, the class was further subdivided bynew classi®cation just within the target class or bydecision rules using auxiliary mapped data notused in the initial strati®cation. This procedurewas repeated until all the classes could be identi®edto a land cover type.

The supplementation of ground survey data withaerial videography was accomplished in southFlorida by the mounting of two 8 mm video recor-ders on a window mount in a light ®xed-wingaircraft. One of the video cameras recorded anapproximately 30630 m swath along the ¯ightline,often providing the resolution needed for specieslevel identi®cation. A wider-angle lens on the othervideo camera recorded a swath of approximately400 m to provide a landscape context to relationshipto the zoomed-in video camera. Slaymaker (1996)provides a detailed description of the con®guration.

A GPS aboard the aircraft and differential post-processing provided georeferenceing for the imageframe that tests suggested was typically within 60 mof the true ground position. Videography was ¯ownin predominately east-west transects approximatelyevery 7�5 min of latitude from the lower Florida Keysto just south of Orlando (see Figure 3 for place loca-tions). When the sequence of transects reachedcentral and north Florida, a Kodak DCS 420 colorinfrared digital camera was substituted for thezoomed-in video camera to improve image reso-lution. The infrared imagery also aided in vegeta-tion species identi®cation. Additional, real-time FMdifferential correction via an Accupoint receiverand a Watson Industries attitude and heading refer-ence system reduced positional errors in the imageframes to 30 m.

Prediction of animal speciesdistributions and species richness

The purpose of the vertebrate species maps was toprovide accurate information of the predicted distri-bution of individual native species in their geo-graphic ranges, and to overlay individual predicteddistributions to produce an overall map of speciesrichness for Florida. Species distributions weremodeled using the Environment Systems ResearchInstitute's Arc/Info software by estimating the geo-graphic range of each species and identifying landcovers suitable for the species' habitat within itsrange. A re®nement of the mammalian speciesmodels was to also estimate the minimum criticalareas of habitat necessary to sustain a viable popu-lation (Allen et al., 2001).

The geographic distribution for mammals wasdetermined by surveying sixteen state and nationalmuseums that included Florida vertebrates, as wellas a review of published sources (e.g., Blair, 1935;Hamilton, 1941; Pournelle, 1950; Sherman 1953;Pearson, 1954; Starner, 1956; Chapman andFeldhamer, 1982; Layne, 1984; Humphrey, 1992).Bird species ranges were based upon the Floridabreeding bird atlas (Kale et al., in press). Reptile andamphibian (herpetofauna) ranges were determinedfrom a statewide occurrence database (Moler, 1999).Butter¯y ranges weredetermined fromOpler (1999).Ant ranges were determined primarily from pub-lished sources, and from the unpublished data ofD. P. Wojcik (1999). Experts of the respectivetaxa reviewed resulting county-level range mapsfor each species.

Animal distribution data for the state of Floridawas almost exclusively at the level of counties. Thus,our original distributions for the modeled specieswere made at the county-level. Using county bound-aries as geographic units for species predictionswould have overestimated distributions of species incases where a species' range extended only partlyinto the county. To reduce this problem, and tofacilitate compatibility with wildlife range cover-ages generated by adjacent states, the county-leveldistributions were joined with a US EnvironmentalProtection Agency hexagon grid system to providedistribution coverage for each species as shown bythese equal-area, 640 ha hexagonal map units.Advantages to using the hexagon grid include itsequal area sampling structure, its independencefrom political and administrative boundaries(resulting in more consistent mapping of animaldistributions), and its hierarchical structure whichcan facilitate increasing or decreasing grid densitiesin future analyses (White et al., 1992).

Species habitat relationships were determinedfrom extensive examination of primary literatureas well as taxa-speci®c treatments and unpublishedreports. These sources are available on the FloridaGap Analysis Project web site: http://www.wec.u¯.edu/coop/Gap. Sources sometimes presented con-tradictory statements concerning wildlife relation-ships of particular species. In these cases we usedthe sources most proximate to or in Florida overother sources, and more recent sources over oldersources. Habitat use data for each species was thenused to build a species-by-habitat matrix for eachtaxa (mammals, birds, herpetofauna, butter¯iesand ants). These matrices recorded predictedthe presence or absence of each species in the71 land cover types that make up the Florida landcover map.

Figure 3. Florida land cover classi®cation. Conservation land boundaries are overlaid as black lines. Place names shownare those places referred to in the text. The State of Florida study area is indicated in yellow on the insert map. Insert mapsource: ESRI Online Data.


In addition, listed species and non-indigenousspecies were assigned codes speci®c to their status.Listed species included species that were `listed' asstate or federally endangered, threatened or speciesof special concern (Florida Game and Fresh WaterFish Commission, 1997). This allowed for compari-son of patterns of species richness among listedspecies, secure native species, and non-indigenousintroduced or invasive species.

Predicted species models that are based primarilyupon habitat (land cover classes) fail to incorporatemany of the basic ecological characteristics of thosespecies. One issue is that home range sizes ofanimals occupying the same landscape may varyby several orders of magnitude. For example,the home range of the golden mouse (Ochrotomysnuttali) in the southeastern United States isapproximately 0�50 ha. In contrast, the home range


of the striped skunk (Mephitis mephitis) mayexceed 300 ha, and the home range of theFlorida panther (Felis concolor coryi) may exceed50 000 ha.

If these species co-occurred in a given favorablelandcoverclasswithanextent (patchsize)of1000 ha,embedded in a matrix of unfavorable habitat, map-ping all three species as present may not re¯ect themore speci®c life history characteristics of each. AFlorida panther could not maintain a viable popula-tion in a habitat patch of 1000 ha, and the arearequirements for a population of striped skunkswould only be marginally met. However, 1000 hawould encompass a viable population of most specieswith small home ranges. On high-resolution maps,commission errorÐthe chance of erroneously in-cluding an animal in a habitat that cannot supportitÐare likely to be high when creating speciesmodels based simply on species-habitat associa-tions. With this in mind, we incorporated informa-tion on the home range of the mammals of Florida toestimate minimum critical areas needed to supportminimum viable populations for each mammal spe-cies (Allen et al., 2001). Incorporating home rangeshould increase the accuracy of species models byreducing the commission error rate.

The home range and dispersal distances of ter-restrial Florida mammals were determined fromextensive literature reviews. We preferentiallyused estimates from studies in Florida, butwhere home range or dispersal estimates speci®cto Florida were not available we used estimates fromnearby locations. Home range estimates were usedto calculate the area required to support a minimumviable population. For our purposes, we crudelyde®ned MVP as being equal to 50 individuals, theestimated minimum number of individuals neces-sary to persist despite demographic stochasticity(Shaffer, 1981). Note that we do not assume that 50is the `real' minimum viable population size formammals, nor that a species' minimum viablepopulation can be precisely de®ned throughout itsrange, rather we chose this as a conservative value.Multiplying home range estimates by 25 calculatedthe minimum critical area required to support aminimum viable population for each species of50 individuals dividing by two. Halving the numberwe multiply home range estimates by accountedfor intersexual overlap among home ranges. Interand intra-sexual home range overlap varies consid-erably among species; we chose complete overlapbetween sexes to produce conservative comparativemodels. No attempt was made to determine min-imum critical area or dispersal distances for batspecies.

Land stewardship

GIS boundaries and attributes for the stewardshipof Florida conservation lands were provided by theFlorida Natural Areas Inventory (FNAI). FNAIidenti®es as conservation lands any property thathas a signi®cant portion of its land area undevel-oped and that has a professional manager or manag-ing agency capable of protecting important elementsof ecological diversity. Additionally, the land willhave a legal mandate to manage and protect impor-tant ecological resources, even if that mandate is notthe primary mission of the agency. Therefore,certain parks such as historical parks that do nothave signi®cant natural areas are not considered aconservation land. Likewise, military installations,which have a primary purpose of national defensebut are also mandated by federal action to protectimportant natural resources, are considered aconservation land.

FNAI staff also produced the GAP protectionstatus rankings for each area in the land steward-ship database following the criteria of Scott andJennings (1994). The four classes in the GAPprotection status rankings are de®ned in Table 1.Information used to develop the GAP protectionstatus was mostly derived from the legal require-ments of different land management categories(e.g., national forests are legally mandated tomanage by multiple use, which lessens the protec-tion a national forest affords biodiversity, versusnational parks are which mandated to protectnatural systems, which ensures that the protectionof biodiversity is given highest priority). FNAIaugmented this approach and also used FNAI'sknowledge of the speci®c management activitieson conservation lands to tailor the protection statusto the current management activities on-site.

Results

Land cover mapping

Table 2 presents the frequency of occurrence of eachland cover type for the state of Florida in squarekilometers, and percent of the state's total arearepresented by the mapped type. The majority ofFlorida is in Mesic-Hydric Pine Forest land cover(18%), forested swamp (14%) or agriculture (23%,including pasture). The agricultural land use is pri-marily converted pinelands. The proportion ofMesic-Hydric Pine Forest that is in plantation farm-ing was not determined. The combined forested

Table 1. Protection status rankings for areas mapped as being managed with conservation objectives. Statuscategories were not applied to all lands of Florida

Status 1An area having permanent protection from conversion of natural land cover and a mandated management planin operation to maintain a natural state within which disturbance events (of natural type, frequency, andintensity) are allowed to proceed without interference or are mimicked through management.

Status 2An area having permanent protection from conversion of natural land cover and a mandated management planin operation to maintain a primarily natural state, but which may receive use or management practices thatdegrade the quality of existing natural communities.

Status 3An area having permanent protection from conversion of natural land cover for the majority of the area, butsubject to extractive uses of either a broad, low-intensity type or localized intense type. It also confers protectionto federally listed endangered and threatened species throughout the area.

Status 4Lack of irrevocable easement or mandate to prevent conversion of natural habitat types to anthropogenichabitat types. Allows for intensive use throughout the tract. Also includes those tracts for which theexistence of such restrictions or suf®cient information to establish a higher status is unknown.


swamp classes are Tropical/subtropical SwampForest, Bay/Gum/Cypress, Loblolly Bay Forest,Swamp Forest, and Cypress Forest. Xeric shrub,Sandhill and sand pine comprise another 4% of thestate, urban classes and freshwater marsh, primar-ily in southern Florida, contribute another 8% eachto the state's land covers. The small percentages ofeach of the individual land cover classes listed inTable 2 are indicative of the heterogeneity ofFlorida's landscape. Because 71 classes are dif®cultto present in a ®gure, Figure 3 is a nineteen-classaggregation of the ®nal land cover classi®cation.Land stewardship boundaries are overlain to illus-trate the patterns and extent of land cover diversityin conservation holdings. Table 3 summarizes theproportion of the state within each of the protectionstatus areas.

As would be expected because of the low pro-portion of the state in status 1 protection, none ofthe state's land covers have a high percentage oftheir area contained within these categories.Status 2 lands, on the other hand, contain a highproportion of the state's mangroves, sawgrassmarsh, and muhly marsh because of their occur-rence in the south Florida everglades parks andpreserves. Sandhill, wiregrass, sand pine, mesic-hydric pine, mixed pine/oak, forested swamplandsof north and central Florida, and the dry prairiesof mostly south central Florida are poorly repre-sented in status 2 lands. These same classes areassociated with the highest concentrations ofspecies richness in Florida. Most of these classesare better represented within protection status3 lands.

Species richness

Species richness for mammals, breeding birds,reptiles and amphibians, butter¯ies, ants and allspecies combined are shown in Figures 4±9.Land stewardship boundaries are overlain on each®gure.

Mammals

The pine/oak and sandhill communities in thepanhandle of Florida potentially support thehighest diversity of mammals (Figure 4). The GAPstatus 3 lands in Eglin Air Force Base, With-lacoochee State Forest and Ocala National Forest,are characteristic of this communities. Sand pine,mesic pine, swamp forest and cypress in the pan-handle through central Florida follow closely behindwith species counts in the low thirties. In broadgeneral terms, potential mammal species richnessvaries along a north±south gradient, and is highestin northern Florida and lowest in southern Florida.This pattern probably is indicative of a decrease inavailable habitat types, rather than a peninsulaeffect.

In southern Florida, the highest species richnessis in southern Florida slash pine, dry prairie, swampforests, and pine rocklands where the maximumnumber of mammalian species is in the lowtwenty's. These areas include lands in and aroundAvon Park Bombing Range, Fakahatchee StrandState Preserve, and lands north of the Caloosa-hatchee River.

Table 2. Land cover frequency

Code Class name Area (sq km) % Total

1 Open water 5675�21 3�222 Tropical hardwood hammock formation 210�38 0�123 Semi-deciduous tropical/subtropical swamp forest 460�57 0�264 Xeric±mesic live oak ecological complex 1363�98 0�775 Mesic±hydric live oak/sabal palm ecological complex 392�30 0�226 Bay/gum/cypress ecological complex 5279�64 3�007 Loblolly bay forest 1393�43 0�798 Cajeput forest compositional group 35�32 0�029 Mixed mangrove forest formation 1085�17 0�62

10 Black mangrove forest 67�91 0�0411 Red mangrove forest 269�78 0�1512 Casuarina forest 4�35 0�0013 South Florida slash pine forest 382�66 0�2214 Sand pine forest 1421�36 0�8115 Xeric±mesic mixed pine/oak forest ecological complex 10 512�39 5�9716 Mesic±hydric pine forest compositional group 30 878�57 17�5317 Swamp forest ecological complex 10 049�82 5�7018 Cypress forest compositional group 6034�24 3�4219 Mixed evergreen Cold-deciduous hardwood forest 5532�42 3�1420 Buttonwood woodland 134�63 0�0821 Mixed mangrove woodland 62�71 0�0422 Black mangrove woodland 11�63 0�0123 Red mangrove woodland 26�18 0�0124 Live oak woodland 1061�02 0�6025 Florida slash pine woodland 496�99 0�2826 Sandhill ecological complex 4697�66 2�6727 Broad-leaved evergreen and mixed evergreen/

cold-deciduous shrubland compositional group921�69 0�52

28 Flooded broad-leaved evergreen shrublandcompositional group

447�25 0�25

29 Dry prairie (Xeric±mesic) ecological complex 1843�19 1�0530 Gallberry/saw palmetto shrubland compositional group 4111�66 2�3331 Brazilian pepper shrubland 89�34 0�0532 Dwarf mangrove ecological complex 676�77 0�3833 Coastal strand 74�09 0�0434 Groundsel-tree/marsh elder tidal shrubland 25�96 0�0135 Xeric scrubland 651�51 0�3736 St Johns wort shrubland compositional group 119�44 0�0737 Saturated-¯ooded cold-deciduous and mixed

evergreen/cold-deciduous shrubland ecological complex3516�91 2�00

38 Saltwort/Glaswort ecological complex 111�37 0�0639 Graminoid dry prairie ecological complex 499�58 0�2840 Sea oats dune grassland 20�38 0�0141 Wiregrass grassland 44�11 0�0342 Graminoid emergent marsh compositional group 2364�61 1�3443 Sawgrass marsh 6101�09 3�4644 Spikerush marsh 200�12 0�1145 Muhly grass marsh 917�68 0�5246 Cattail marsh compositional group 253�36 0�1447 Salt marsh ecological complex 196�83 0�1148 Sand cordgrass grassland 136�38 0�0849 Black needle rush marsh 717�06 0�4150 Saltmarsh cordgrass marsh 613�91 0�3551 Saltmeadow cordgrass/salt grass salt marsh 1�78 0�0052 Sparsely wooded wet prairie compositional group 89�02 0�0553 Dwarf cypress prairie 697�71 0�4054 Temperate wet prairie 633�11 0�3655 Maidencane marsh 194�92 0�1156 Forb emergent marsh 1571�51 0�8957 Water lily or ¯oating leaved vegetation 626�15 0�3658 Periphyton 0�00 0�00

(Continued on next page)


Code Class name Area (sq km) % Total

59 Sand, beach 256�32 0�1560 Bare soil/clearcut 4721�47 2�6861 Pavement, roadside 390�97 0�2262 Urban 4088�33 2�3263 Urban residential 6709�47 3�8164 Urban open/others 1904�26 1�0865 Agriculture 20 950�22 11�8966 Pasture/grassland/agriculture 14 882�27 8�4567 Agriculture/groves/ornamental 4238�82 2�4168 Agriculture/con®ned feeding operation 271�86 0�1569 Extractive 1127�16 0�6470 Recreation 434�65 0�2571 Cloud 202�95 0�12

Total 176 187�56

Table 2. Continued

Figure 4. Mammalian species richness. Conservation land boundaries are overlaid as black lines.


Mammal species richness is high throughoutnorth Florida and the panhandle. In particular,unprotected areas of high species richness includeboth coasts of north Florida and the panhandle,

the lands between Eglin Air Force Base and Black-water River State Forest, and the lands betweenOsceola National Forest, Camp Blanding MilitaryReservation, and Ocala State Forest.

Birds


Median bird diversity as modeled is between 52 and56 species. Bird species richness is highest in swampforests, pasture, agriculture and urban classes.Because our interest is primarily in species richness

Table 3. Summary by land stewardshipstatus area. Status categories are de®ned inTable 1

Area (sq km) % of State

Status 1 314�23 0�19Status 2 16 609�87 9�79Status 3 21 314�42 12�57Status 4 131 349�29 77�45Total 169 587�81 100�00

Figure 5. Avian species richness excluding urban and cropblack lines.

within natural areas, Figure 5 illustrates birdspecies richness when urban and agriculture otherthan pasture/grassland are excluded. All forestedclasses are associated with high counts of birdspecies though the swamp forest classes (includingcypress and bay) are signi®cantly higher than otherforest land covers. As with the mammals, there isevidence of decreasing richness along a north±southgradient. An exception to this gradient is an areaof high avian richness in extreme southwesternFlorida.

State and federal lands such as Eglin Air ForceBase, Withlacoochee State Forest, Osceola NationalForest, Apalachicola National Forest, FakahatcheeStrand State Park, and Big Cypress NationalPreserve provide habitat for a high diversity ofavian species. Unprotected areas of avian diversity,

land covers. Conservation land boundaries are overlaid as


in order of species richness, exist along both coastsof north Florida, east central Florida, and thepasture lands of the Immokalee Rise (southwestFlorida) and north of the Caloosahatchee River.

Reptiles and amphibians

The highest modeled richness of reptiles and amphi-bians is associated with open water, swamp forests,and sandhill land covers. Species diversity forhepetofauna is highest in the panhandle and decrea-ses in central and southern Florida, mirroring thebroad pattern displayed by the other vertebratetaxa. High richness habitats in central to northernFlorida support 40 to species. In southern Florida,the maximum species richness is in the mid-thirties.

Figure 6. Reptile and amphibian species richness excludingare overlaid as black lines.

Figure 6 presents reptile and amphibian speciesrichness for all land covers excluding anthropo-morphic (crops and urban) development.

Eglin Air Force Base, Apalachicola NationalForest, and the Big Bend Wildlife ManagementArea are examples of some of the highest reptileand amphibian diversity in protected areas. Thebottomland and wet forested areas of the Gulf Coastappear to provide the best opportunities for addi-tional protection of species richness.

Ants

Across the Florida peninsula, potential ant speciesrichness, like vertebrates, displays a pattern ofhighest richness to the northern and decreasing

urban and crop land covers. Conservation land boundaries


richness towards the southern peninsula. This isprobably due to decreasing habitat diversity and anincreasing prevalence of saturated and inundatedhabitat types. Species composition also changesacross this gradient, with an increasing prevalenceof species of West Indian origin in southern Floridaconcomitant with a decreasing important of speciesof Northern American origin.

Modeled ant species richness in southern Floridais highest in the pine rocklands, southern Floridaslash pine, and tropical hardwoods. Pine commu-nities continue to be an important habitat for antsin central and northern Florida. The sand pine andpine/oak land covers in northern Florida are suit-able habitat for the highest numbers of ant speciesin the state (Figure 7).

The highest modeled ant species richness iswithin the xeric central Florida sandhill and sand

Figure 7. Ant species richness. Conservation land boundar

pine communities protected by Ocala NationalForest and Withlacoochee State Forest. The highestunprotected diversity is in the forested land coversof north Florida. In particular, the lands betweenand adjacent to Ocala National Forest, CampBlanding Military Reservation and Paynes PrairieState Preserve.

Butter¯ies and skippers

Butter¯ies and skippers are the only group modeledthat have the highest diversity in southern Florida.The Everglades and Big Cypress National Preserveare suitable habitat for 40 to 50 butter¯y species.Everglade's marsh north of Timiami Trail hasspecies richness counts as high as 56. Maximumspecies richness in northern Florida is in the teens,except for graminiod marsh land covers where

ies are overlaid as black lines.

Figure 8. Butter¯y and skipper species richness. Conservation land boundaries are overlaid as black lines.


the number of butter¯y species can be in thethirties and even the sixties in restricted smallareas (Figure 8).

At the state level, areas of high butter¯y diversityall appear to be within the con®nes of conservationlands. The largest areal extent is in south Floridawith smaller pockets of high species richness incentral Florida and at the mouth of the ApalachicolaRiver.

All species combined

Total species richness (mammals, birds, reptiles,amphibians, ants, and butter¯ies combined) is pre-sented in Figure 9. Overall richness of all taxa

mapped follows a north±south gradient, withhighest richness in the north and decreasing rich-ness southward. Highest species richness is associ-ated with swamp forest and sandhill land covers. Insouthern Florida, pine communities provide habitatfor the largest number of species. Species richnessoverall follows the pattern of most of the individualgroups, with highest diversity in the panhandle ofFlorida.

Status 1 conservation lands have a very highdiversity of species, but because of their smallextent, the species all represent a tiny proportionof the state's distribution. The occurrence of specieswithin the differently ranked conservation landsclosely re¯ects the relative proportion of the statuscategory in the state.

Figure 9. Total Species richness. Conservation land boundaries are overlaid as black lines.


Discussion

The FL-GAP project provides a tool for settingpriorities for biodiversity conservation. This typeof information can be used to assist in prioritizationof land acquisition, restoration, and managementactions as well as evaluating the potential effectsof development and other changes in land use.Application is most relevant at regional levelsgiven the broad scale patterns of vegetation andpotential habitat distributions for a large number ofspecies the technique employs. The land coverclassi®cation provides a regional perspective ofland cover patterns, juxtapositions and occurrencesin the landscape. Use in smaller scale applicationsshould be limited to analyses of parcel contextwithin the regional landscape. We recommendthat application in any project smaller than

regional scale be carefully reviewed and revisedappropriate to the scale required for the objectivesof the project.

Because all of the terrestrial vertebrates of thestate have been mapped along with two invertebrategroups, these products offer natural resource man-agers a tool for multi-species protection relevant toallocation of land resources. Potential habitats inthe spatial databases can be used to distinguishspecies communities by identifying species sharingthe same habitat requirements or species sharing acommon area, perhaps within a matrix of diversehabitats. The US Fish and Wildlife Service in southFlorida, for example, is using the FL-GAP models tohelp answer questions about how protection ofpotential habitat for individual T&E species cancontribute to the protection of potential habitat forall other terrestrial vertebrates.


While these models are not intended for evalu-ation within small, site-speci®c projects, intelligentecological restoration and permitting requiresspatially-explicit knowledge of conditions proximateand regional to project sites. Species often aredependent on habitat areas larger than thoseimpacted by a single project (Gosselink et al., 1990)and the spatial patterns of habitat adjacent to aspeci®c site may as important as the within-sitehabitat (Pearlstineetal., 1997;Saundersetal., 1991).Examples include determination of areas suitablefor viable and sustainable populations (habitat andrisk assessment), areas of socioeconomic and envi-ronmental con¯ict, and optimization of develop-ment footprints to protect natural systems.

An alternative to the species-habitat models usedin FL-GAP are occurrence models that selecthabitat based on areas of known species occurrence.These two forms of habitat modeling are compli-mentary. When results of the two approaches arecompared, areas of overlap provide robustness tothe analysis, while areas of little or no overlap willindicate limitations in the data layers or areaswhere additional information is needed to bettermodel the species. Areas of the species-habitatmodels that do not overlap with occurrence datamay also draw attention to potential areas forspecies reestablishment. In Florida, both theFlorida Fish and Wildlife Conservation Commissionand The US Fish and Wildlife Service are involvedin species reestablishment activities.

We expect that the FL-GAP models overestimatethe spatial extent of most species' distributions.The vertebrate, butter¯y, and ant habitat-af®nitymodels rely on associating the potential for aspecies' presence with speci®c land covers. Themammal models are re®ned to some extent by con-sideration of minimum contiguous area to maintaina viable population. Regardless, there are manyfactors in habitat selection that have purposefullynot been considered because of resource and timelimitations, the availability of statewide data, and/orinadequate knowledge of species response to theenvironmental variables. Reptiles and amphibiansare an example of a group that may be responding tosoil type, litter accumulation, moisture conditionsand proximity to streams rather than directlyto vegetation composition. Other parameters canreadily be identi®ed for other species or speciesgroups including land cover structure (such as age,height and layering of vegetation, presence ofsnags), and the juxtaposition of desirable vegetationcover to other desirable cover types (e.g., foragingversus nesting) or undesirable features (e.g., roadsor heavy recreational activity).

Conclusions

The Fl-Gap mapping of Florida species richnessclearly suggest a high diversity of most of the taxa innorth Florida and in particular, the panhandle ofthe state. Swamp, mixed pine/oak, sandhill scruband longleaf pine, and ¯atwoods pine all contributeto the wildlife species richness of this area of theState. Additionally, for many taxa this region is anarea of overlap of species with northern af®nitiesand species with southern af®nities. The release of®ndings from a joint project of The Nature Conser-vancy and the Association of Biodiversity Informa-tion, `Precious Heritage: The Status of Biodiversityin the United States' (Stein et al., 2000) corro-borates the importance of Florida panhandle bio-diversity using a different approach to modelingdiversity. The Precious Heritage project uses stateHeritage Program data to map species richness of anarea weighted by relative rarity of the species asmeasured by how restricted its distribution is. Theanalysis pinpoints the panhandle of Florida as oneof the six most signi®cant areas of biodiversity inthe United States.

In southern Florida, with the exception of butter-¯ies and skippers, forested classes (swamp and pinerockland) appear to support the highest diversity. Incentral Florida, species-rich classes include swamp,pine ¯atwoods, xeric scrub, and sandhills. It isimportant to note that xeric scrub in the southcentral Florida region is underrepresented in theland cover classi®cation and resulting in lowerspecies richness in much of this area than wouldotherwise be the case.

The pine ¯atwood, xeric pine, and xeric scrubcommunities in Florida are major contributors tothe species diversity of the state and are some of themost threatened because of changing ®re regimesand their suitability for development or slash pinefarming. The unique pine rocklands of south Floridaare threatened by exotic invasive vegetation. Innorthern Florida, the sandhill scrub and longleafpine habitats of Eglin Air Force Base are protectedby an aggressive natural resource managementplan that includes continuing research, adaptivemanagement, and longleaf pine ecosystem restor-ation. Ocala National Forest and WithlacoocheeState Forest in central Florida are two additionallarge areas of xeric community conservation. Inother areas of north and central Florida loss of thesevaluable upland habitats is progressing steadily.

The Fl-Gap biodiversity project has compiled abaseline for predicting species distributions andrichness statewide that must now be re®ned with


studies scaling from landscape analyses to on-the-ground validation. Landscape context re®ne-ments of the species models should address issues ofsustainability and resilience. Key questions thatneed to be answered include: (1) How much con-tiguous and non-contiguous habitat is required tosustain wildlife populations into the future? (2) Howwill con®gurations of landscape elements maintainhabitat in dynamic systems impacted by naturalprocesses such as ®re and hurricane disturbance?(3) What factors in¯uence the process of animalsmoving among landscape elements? (4) How doesthe proximity of different land covers change theviability of selected habitats? (5) How do results ofthis analysis change at different scales, includinggrain (the minimum unit of measurement) andextent (the size of the study area and/or length oftime over which it is studied)? (6) Finally, it iscritical to incorporate the temporal change in thelandscape for the applications to be meaningful. Ifwe stop at conservation of current known `hot spots'of species richness, we risk great loss as dynamics inthe environment shift ecological conditions.

Fl-Gap is expected to be an on-going project. Thisreport is only the ®rst iteration in a process oflearning to adapt and use a new, ecologically-basedNational Vegetation Classi®cation Scheme tostatewide mapping and apply those results aswell as other appropriate mapped data sources toa measure of Florida's biodiversity and potentialsfor conservation.

Future iterations of GAP biodiversity measureswill increasingly incorporate parameters of biologic-al integrity and a wider range of species taxa. Theexperiences gained and data compiled by this andother states engaged in the US National GapAnalysis Program, increasing availability of geo-referenced, broad extent mapped biological andgeological data, improved earth-sensing techno-logies, and the dissemination of better approachesfor landscape modeling and statistical analysiswill continue to strengthen the scienti®c basis fornatural resource decisions in the state of Florida.

Acknowledgments

This study was funded in part by the US GeologicalSurvey, The Florida Fish and Wildlife ConservationCommission, Eglin Air Force Base, DOD, and theFlorida Agricultural Experiment Station journal seriesmanuscript number R-08162.

Signi®cant contributions of time were alsomade by Florida Natural Areas Inventory, CoastalServices Center, NOAA, Center for Coastal Geology,

USGS, and many individuals with particular thanksto Nick Ansay, Joe Aufmuth, Valentina Boycheva,Michael Cook, Steve Cox, Michael Frankenberger,Karen Garren, E. Paul Gonzalez, Lisa Ojanen, andRoger Winstead.

References

Allen, C. R., Pearlstine, L. G. and Kitchens, W. M. (2001).Modeling viable mammal populations in gap analyses.Biological Conservation 99, 135±144.

Blair, F. W. (1935). Some mammals of southern Florida.American Midland Naturalist 16, 801±804.

Bureau of Census (1960). Census 1960: Summary ofFile 1: 100 Percent Data. Department of Commerce,Viewed on 20 November 2001.

Bureau of Census (1990). Census 1990: Summary ofFile 1: 100 Percent Data. Department of Commerce,Viewed on 20 November 2001.

Bureau of Census (2000). Census 2000: Summary of File1: 100 Percent Data. Department of Commerce,Viewed on 20 November 2001.

Chapman, J. A. and Feldhamer, G. A. (1982). Wildmammals of North America. Baltimore, MD: TheJohns Hopkins University Press.

Cox, J., Kautz, R., MacLaughlin, M. and Gilbert, T.(1994). Closing The Gaps In Florida's Wildlife HabitatConservation System. Tallahassee, FL: Florida Gameand Fresh Water Fish Commission, Of®ce of Environ-mental Services.

Crist, E. and Cicone, R. (1984). Applications of thetasseled cap concept to simulated Thematic Mapperdata, Photogrammetric Engineering and RemoteSensing, 50, 343±352.

ERDAS (1999). ERDAS Field Guide. 5th Edition.ERDAS, Inc. Atlanta, GA.

Florida Game and Freshwater Fish Commission (1997).Of®cial list of endangered and potentially endangeredfauna and ¯ora in Florida. Tallahassee (FL): FloridaGame and Freshwater Fish Commission.

Gosselink, J. G., Lee, L. C. and Muir, T. A. (eds) (1990).Ecological Processes and Cumulative Impacts:Illustrated by Bottomland Hardwood WetlandEcosystems. Chelsea, MI: Lewis Publishers, Inc.

Grossman, D., Goodin, K. L., Li, X., Wisnewski, C.,Faber-Langendoen, D., Anderson, M., Sneddon, L.,Allard, D., Gallyoun, M. and Weakley, A. (1994).Standardized national vegetation classi®cationsystem. Report by The Nature Conservancy andEnvironmental Systems Research Institute for theNBS/NPS Vegetation Mapping Program. Denver,Colorado: National Biological Service.

Hamilton, W. J. (1941). Notes on some mammals of LeeCounty, Florida. American Midland Naturalist 25,686±691.

Humphrey, S. R. (ed) (1992). Rare and Endangered Biotaof Florida: Mammals. Gainesville, FL: University ofFlorida Press.

Jennings, M. D. (1993). Natural terrestrial cover classi-®cation: Assumptions and de®nitions. Gap AnalysisTechnical Bulletin 2. Idaho Cooperative Fish andWildlifeResearchUnit,Moscow,ID:UniversityofIdaho.


Jenson, J. R. (1996). Introductory Digital Image Process-ing: A Remote Sensing Perspective, Prentice Hall,Upper Saddle River, NJ.

Kale, H., Pranty, B., Stith, B. and Biggs, C. (1999). Atlasof Breeding Birds of Florida. The Florida AudubonSociety.

Layne, J. (1984). The land mammals of South Florida. In:Environments of South Florida, Past and Present II,(Gleason, P. J. ed.), pp. 269±295. Coral Gables, FL:Geological Society.

Moler, P. (1999). Statewide Occurrence Database,Florida Game and Fresh Water Fish Commis-sion, www.npwrc.usgs.gov/resource/distr/lepidb¯yusa/bl¯yusa.htm.

Opler, P. (1999) Personal communication.Pearlstine, L., Brandt, L. A., Mazzotti, F. J., Kitchens,

W. M. (1997). Fragmentation of pine ¯atwood andmarsh communities converted for ranching and citrus.Landscape and Urban Planning 38, 159±169.

Pearlstine, L. and McKerrow, A. (1999). CompositionalGroups and Ecological Complexes: A Method forAlliance-Based Vegetation Mapping. In Gap AnalysisBulletin (E. Brackney and M. Jennings, eds), v. 7,pp. 16±17. Moscow, Idaho: U.S. Geological Survey,Biological Resources Divsion.

Pearson, P. G. (1954). Mammals of Gulf Hammock, LevyCounty, Florida, American Midland Naturalist 51,468±480.

Pournelle, G. H. (1950). Mammals of a north Floridaswamp. Journal of Mammalogy 31, 310±319.

Saunders, D. A., Hobbs, R. J. and Margules, C. R. (1991).Biological consequences of ecosystem fragmentation: areview. Conservation Biology 5, 1±32.

Scott, J. M., Csuti, B., Jacobi, J. D. and Estes, J. E.(1987). Species richness. Bioscience 37, 782±788.

Scott, J. M. and Jennings, M. D. (eds) (1994). A hand-book for Gap Analysis. Moscow, Idaho: Idaho Coopera-

tive Fish and Wildlife Research Unit, University ofIdaho.

Scott, J. M., Davis, F., Csuti, B., Noss, R., Butter®eld, B.,Groves, C., Anderson, H., Caicco, S., D'Erchia, F.,Edwards, Jr., T. C., Ulliman, T. and Wright, G. (1993).Gap analysis: A geographic approach to protection ofbiological diversity. Wildlife Monographs 123.

Shaffer, M. L. (1981). Minimum population sizes forspecies conservation. BioScience 31, 131±134.

Sherman, H. B. (1952). The list and bibliography of themammals of Florida, living and extinct. QuarterlyJournal of the Florida Academy of Sciences 15,100±126.

Slaymaker, D. M., Jones, K. M. L., Grif®n, C. R. andFinn, J. T. (1996). Mapping deciduous forest inSouthern New England using aerial videography andhyperclustered multi-temporal Landsat TM imagery.In Gap Analysis, A Landscape Approach to Biodiver-sity Planning, (J. M. Scott, Tear, T. H. and Davis, F. W.eds). pp. 87±101. Bethesda, MD: American Society ofPhotogrammetry and Remote Sensing.

Starner, B. A. (1956). Notes on the mammals in threehabitats in North Florida. Quarterly Journal of theFlorida Academy of Sciences 19, 153±156.

Stein, B., Kutner, L. and Adams, J. (eds). (2000). PreciousHeritage, The Status of Biodiversity in the UnitedStates. Oxford University Press.

The Nature Conservancy. (1997). International Classi®-cation of Ecological Communities, Terrestrial Vegeta-tion of The Southeastern United States. Chapel Hill,NC: The Nature Conservancy.

White, D., Kimerling, J. and Overton, S. (1992). Carto-graphic and geometric components of a global samplingdesign for environmental monitoring. Cartography andGeographic Information Systems 19, 5±21.

Wojcik, D. (1999). U.S. Department of Agriculture,Gainesville Florida, unpublished data.

assessing state-wide biodiversity in the florida gap ... · assessing state-wide biodiversity in...

Documents