sign in sign up
<<
Home , Everglades National Park

PEER.REVIEWED ARIICTE

GISDatabase Development forSouth 's NationalPafts and Preserves

R. Welch, M. Remillard,and R.F. Doren

Abstract Introduction The University of Georgia's Center for Remote Sensing and The Center for RemoteSensing and Mapping Science(cnvs) Mapping Science (CRMS)is working in conjunction with the at The University of Georgiais working with the U.S. Depart- U.S. Department of Interior's Nationol Park Service (NpS)to ment of Interior's Service (Npslto utilize a construct a geographic information system (cts) database combination of satellite imaging, aerial photographic,Global and associated detailed vegetation maps for the Positioning System (cps), and geographicinformation system National Park, , Big Cypress National (cIs) technologies to develop a databasein GISformat for Preserve,and the Flofida Panther Refuge. Preservation of over one million hectares (ha) of ecologically unique Ever- these wetland areas, threatened by urban ex- glades wetlands in south Florida. Although the southern tip pansion, nutrient runoff from agricultural \ands, encroach- of Florida was the entry point for early explorers into the ment of exotic plant species,and increasedrecrcational use, New World, the Evergladesremain one of the last portions of is a topic of national concern, Development of the database the United Statesto be accuratelymapped at any level of de- and maps is made possible by the integration of Global Posi- tail. tioning System (cesJ, satellite remote sensing, air photo inter- This vast, flat terrain, wetland study area includes Ever- pretation, and helicopter-assistedfield verification gladesNational Park, BiscayneNationai Park, Big Cypress procedures. A digital satellite image mosaic prepared from ,and the Refuge(Figure 1). eight sror panchromatic images of tO-m resolution and geo- Collectively, this area will be referred to as the "Parks." It coded to ground control points in the urM (Nto as) coordi- extends roughly from on the east to Naples on the nate system to an accuracy of t1 to 1-1.5 pixels forms the west, southward to , and representsthe remain- coordinate reference layer for the GIS database. Vegetation ing lands of the greater Big Cypress and Everglades patterns and, where appropriate, hydrographic and transpor- ecosystemsthat once covered approximately one-third of the tation features are digitized directly from 4x paper print en- Florida peninsula (Light and Dineen, 199a).The growth of largements of National Aerial Photogrophy Program (N,spp) the Miami urban area, the encroachment of exotic plaats color infrared aerial photographs recorded in L994 and lgg7. such as Melaleuca quinquenervio (cajeput)and Scftrnusfere- These digitized vector files are rectified to ground control binthifolius (Brazilian pepper), the expansion of agricultural transferred from the rectified SpoT satellite images. Planimet- Iand along the margins of the Parks,and increasedrecrea- ilc enors generclly are less than + 10 m. The vegetation lay- tional use combine to threaten the flora and fauna in this re- ers in the GtSdatabase are classified according to a prcto- gion (Dueveret al., L986;Doren ef o1.,tggo; Davis and type Everglades Vegetation Classification System being Ogden,1994). developed by NPS,CRMS, and South Florida Water Manage- Although portions of the Parks have been mapped for ment District (SFwMD)personnel. Ground truth collection and particular researchinterests, neither detailed maps nor a verification of the thematic accuracy of interpreted vegeta- comprehensivedatabase of Evergladesvegetation exists at tion polygons and boundaries is facilitated through the aid this time (e.g.,McPherson, 1973; Gunderson and Loope, of the SPor satellite image mosaic, and a laptop computer 1982;Olmstead et a1.,1983;Rose and Draughn,1991). The inte$aced to a GPSreceiver mounted in a helicopter. The South Florida Water ManagementDistrict (SFWMD)recently flight track of the helicopter is displayed in real time on the used satellite image data to develop a land-covermap of SPOTimage mosaic, and, as required, attribute information is SFWMDWater ConservationArea 2A. Iocated northeastof the entered into the computer. The digital GISdatabase and EvergladesNational Park (Rutcheyand Vilchek, 1994; fensen 1:24,000-scolevegetation maps will provide the NpSwith the et al., l,ssi). Rutchey and Vilchek (t99+) concluded, how- detailed, up-to-date spatial information needed to manage ever, that, due to the diverse nature of Evergladesvegetation, the Parks of south Fhorida. the accuracyof thematic classificationresults may be im- proved by using low altitude aircraft multispectral scanner data and aerial photographs. The SFWMDis now cooperating with the NPSto coordinate and develop joint mapping proce-

R. Welch and M, Remillard are with the Center for Remote Sensingand Mapping Science(CRMS), Department of Geog- PhotogrammetricEngineering & RemoteSensing, raphy, The University of Georgia,Athens, GA 30602-2503. Vol. 0r, No. 11, November1995, pp. 1,371,-1.381..

R.F. Doren is with the South Florida Natural ResourceCenter 0099-11 1 2/95/61 11-1 3 71$3.00/0 (SFNRC),Everglades National Park, 40001 StateRoad 9336, 1995 American Society for Photogrammetry Homestead,FL 33034-6733, and Remote Sensing

PE&RS PEER.REVIEWED ARIICLE

time collection of ground truth information by helicopter surveys.The surveying,remote sensing, and bls procedures being used to build the databaseare describedin this paper. DataSources Data sourcesfor the proiect include maps produced by the U. S. GeologicalSurvey (uscs) and the Nps South Florida Natural ResourceCenter (SFNRC),satellite image data, and color infrared (cn) aerial photographs.The majority of the USGS1:24,000-scale quadrangles of the Parks were produced from aerial photographsin the 1960sand early 1970sas mul- ticolor orthophotomaps(Pumpelly, 1967).These orthophoto- maps have extensivecartographic treatment such as color patterns,lettering, and symbols placed on the photo image which tend to mask vegetationpatterns. Although statements on these orthophotomapsindicate that they comply with U.S. National Map Accuracy Standards(Nuas) (* z.z RMSE*' for well-defined features),many of the map sheetsfor the Ev- ergladesNational Park and Big CypressNitional Preserve show extensivewetland areaswith no "well-deffned" fea- tures. Thus, for the interior of the study area,the USGSortho- photomaps are dated, provide limited detail, and are of questionablegeometric reliability. By contrast,the standard 1:24,000-scaleUSGS topographic line maps for the developed R'i':J;*'"i:.*" jfl,?l'*' areassurrounding the Parks (revisedin the 1980s)are a valu- o 2s 50h 3 f,'3,5J'ffil*"iHri** able (Ccrs) D. 8iryne Natoml Pa* sourcefor ground control points on the perimeter SCle of the study area. Largescale vegetation maps (t:t+,000 to 1:43,000scale) Figure1. TheSouth Florida National Parks and Preserves produced by the SFNRCin the 1970sand early 1980sprovide studyarea is shownby a dashedline, and the sPoTim- detailed vegetationinformation for limited areasof the Ever- agesemployed to createa geccodedmosaic are indi- gladesNational Park and Big CypressNational Preserve.Al- catedbv solidlines. though these maps are useful for confirming the identifica- tion of vegetationin a few specific locations,they only cover a small fraction of the study area and do not conform to spe- cific geometricaccuracy standards. dures and vegetation classifications to create a vegetation da- Existing map coverageof the Parks,therefore, was inade- tabase for the Parks water conservation and areas. quate for deriving a cCPnetwork of sufficient density and ac- With limited road access to tracts of the study extensive curacy to createa GtSdatabase directly from available CtR area and overland travel made difficult by wetlands, exposed aerial photographs.Alternative sourcesof control, namely pinnacle rock, and entangled forest, it was deter- GPS,aerotriangulation, and satellite image data, were consid- mined that cPS surveys in conjunction with satellite image ered. Aerotriangulation was ruled out becauseof the charac- data would be required to bridge remote and pro- the areas teristics of the National Aeronautics and SpaceAdministra- vide an accurate control network for cIs database develoo- tion (Nasa) aerial photographs(described 6elow) available at ment (Welch et al.,'tgg2; Welch and Remillard, 199a). Aiso, the outset of the project and the difficulties in establishinga it was concluded that detailed mapping of vegetation, hy- control network for severalhundred photographs.Satellite drography, and transportation features should be undertaken images,however, in combination with a cps survey and with aerial photographs. Consequently, the objectives of this USGS1:24,Ooo-scale topographic line maps for the perimeter project were defined as follows: of the study area,offered a rapid and economical means of . Constructa geocodedsatellite image mosaic of the Parksthat establishinga densenetwork of ccps of sufficient planimetric will provide the referencelayer for a cts database; accuracyto control the aerial photographs.Although the use o Producea detailed cls databasein digital format that con- of satellite imagesto provide control for aerial photographs forms to an accuracystandard of approximately + 19 trt is contrary to traditional practice, past experiencehas shown (RMSE.") root-mean-squareerror for well-defined planimetric that, becauseof their excellent internal geometry,only four features,and includes the distribution of plant communities to six GCPsper image are required to geocodea SPOTsatellite within the Parks;and +0.5 +1.0 . Assessvegetation damage caused by . imageto a planimetricaccuracy of to pixel- or + 5 to + 10 m on the ground (Welch,1985; Welch ef o/., In order to meet these objectives, it was necessary to ac- 1985). Such accuraciesare compatible with NMASfor '1.:24,00D-scale curately geocode satellite images and develop techniques for maps and were consideredacceptable for using control transferred from the satellite images to rectify establishinga geoioded image database. several hundred aerial photographs; establish efficient photo- Satellite imagesrecorded by three different sensorsys- interpretation and feature encoding procedures compatible tems were consideredas potential sourcematerial for a geo- with a new classification system that takes into account vege- coded image database:(1) spor panchromaticimages of 10-m tation species, human impacts, and hurricane damage; and pixel resolution, (2) RussianKATE-2oo images of 12-m pixel integrate cPS surveys with attribute recording and digital im- resolution (Spradley,1994), and (3) Landsat Thematic Map- age processing on a laptop computer to facilitate the real- per (rtr,t)images of 28.5-m pixel resolution (Table LA; Figure

L372 PE&RS PEER.REVIEIYED ARTICTE

Figure2. A road intersection(used as a controlpoint) is visi- ble on three types of satelliteimage data: (a) sPor panchro maticimage of 10-mpixel resolution, (b) RussianKATE-200 ima1eof I2-m pixelresolution, and (c) LandsatThematic Mapper(rM) imageof 28.5-m pixel resolution.Because of their superiorspatial resolution,seor imageswere identified as the primarysatellite image data set for derivinga control network.

2). Becauseof their recent date of acquisition (r00a), supe- after Hurricane Andrew (Table 1B). Although the timing of rior 10-m spatial resolution, and availability from the NPS these photographsis excellent for hurricane damageassess- and Spwun, the SPOTimages were identified as the primary ment, the 1:65,000-scalephotos (23 by 23 cm) lack sufEcient satellite image data set (Figure 3). The RussianKATE-200 im- detail for vegetationstudies. The 1:32,500-scale(23- by a6- age covered the entire study area, but poor contrast limited cm) photos, on the other hand, depict adequatedetail, but its use. Both the KATE-200and multispectral Landsat TM im- their use is complicated by the large, rectangular format and ageswere used to plan the locations of cps control points the non-standard, cross-track (instead of along-track) orienta- and to locate supplementalcontrol points neededto geocode tion of the long axis (aO cm) of the camera focal plane dur- the SPOTimages. ing photo acquisition. For example, it is difficult to locate Aerial photographs were acquired with mapping cameras any commercial company or civilian government agency that (23- by 23-cm format) and frame reconnaissancecameras (23- can (or will) produce full format photographic enlargements by 46-cm format) from an altitude of tg,aoo m (65,000ft) by from 23- by 46-cm fi.lm transparencies,and it is exceedingly the NASAU-2 researchaircraft on 5 April 1990 and on 30 No- cumbersometo work with stereomodels having dimensions vember and 3 December1992, approximately three months of re by 46 cm.

PE&RS PEER.REVIEWED ARTICTE

TaeLe1. l. Snrcu-rreluaee Dnrnron Sourx Flonrol

Satellite Sensor Pixel Size Spectral Bands Areal Coverage No. of ScenesRequired Landsat TM 28.5m Multispectral (z bands) 185by 185km 2 SPOT HRV 10m Panchromatic (1 band) 60 by 60 km 8 KATE-2OO !2m Panchromatic (1 band) 192by 228km I B. RecentAerial PhotographicCoverage of South Florida

Date Format Scale Film Type Camera Focal Length Agency 05 Apr 90 23 by 23 cm 1:65,O0O so-131(cIR) Wild (Leica] 304.97mm NASA RC-10 (12 in.) 05 Apr 90 23 by 46 cm 1:32,500 so-131(cIR) Hycon 609.6mm NASA HR 732 (2a in.) 30 Nov 92 23 by 23 cm 1:65,000 so-131(crR) Wild (Leica) 304.97mm NASA and RC-10 (12 in.) 03 Dec 92

30 Nov 92 23 by 46 cm 1:32,500 so-131(cIR) Hycon 609.6mm and HR 732 (24 irL.') 03 Dec 92

15 Mar 94 23 by 23 cm 1:40,000 so-134(crR) Zeiss 153.224mn USGS and RMK A 15/23 (6 in.) NAPP Mar 95

During 1994 and 1995, as part of the USGSNational Aer- vege-tation,transportation, and hydrographic features in a cls ial Photography Program (NaneJ,1:40,000-scale CrR aerial database,it was first necessarv to esttblish a network of ccps photographs of the study area were recorded with standard distributed throughout the study area adequate for rectifying ?3- by 23-cm format mapping cameras. The NApp photos of- the eight SpoT images. The survey to establish GCpswas in-" fer severaladvantages over the NASAphotos: (1) superior fluenced by the existing road network and by the availability quality and color conbast, (2) standard 23- bv 23-cm format, (3) stereo coverageregistered to the 7.5-minuie topographic quadrangles, and (4) a cunent record of vegetation patterns in south Florida (Plate 1). These photos are being used as the primary remote sensing data set for mapping vegetation. The 1_990and 1992 NASAphotographs are available for assessing the damage caused by Hurricane Andrew, Methodolog! In order to meet the initial objectives of providing a geo- coded satellite image mosaic and to ensure the registration of

Aerial Photographs

SPOTlmage llosaic

UTMCoordinate Rsforence System

Figure3. Remotesensing data for the project includedcolor infraredaerial photographs Plate 1. Sample1994 NAppcolor infrared(ctR) aerial pho- and sPoTpanchromatic images of 1Gm reso- tographof mangroveforest locatedon the west coast of lution.The spoTimages form a geocodedmo- EvergladesNational Park. Note the contrastbetween saic tied to the NorthAmerican Datum of healthymangrove (red) and damagedmangrove forest 1983 (NAD83). (green)caused by HurricaneAndrew. PEER.REVIEWED ARIICIE

of 1:24,000-scale topographic line maps of recent vintage for study area in order to provide the interior control necessary the lands adjoining the Parks. Thus, it was decided to use to rectify the satellite images. the existing maps for perimeter control and to conduct the The ground survey was conducted by Global Satellite cPs suruey along the few roads through the middle of the Surveys,becatur, Alabama, with assistancefrom cnus and Nes perso.rnel, Four Trimble Navigation 400osELand Sur- veyoi receivers were employed using static differential cps suiveying techniquesto establisha network of-23-image- identifiable ccPs along 300 km of roads through the study area (Figure 4). These control points included the intersec- tions of roads and of roads and , both of which are easilv located on the SPOT10-m panchromatic images. In ad- dition, the survey included eight monumented National Geo- detic Survey (Ncs) and Florida GPS(FLGPS) base points of - Order B (1:i,000,000)accuracy (or better) which were used to adjust the network. Basedon checks conducted in the post-processing-to adjustment of the GPSobseryations, the accu- iu"y which the Universal Transverse Mercator (urvr) coordinatesof the 23 points were establishedas referenced to the North American Datum of tgae (Nan as) was about + 0.05 m (rutsn*r).

Geocodedlmage Mosalc Initially, consideration was given to using a block adiustment approath to create a seamlessmosaic (Lenzen and Foresman, f'S'9s).However, this approach necessitatedthe use of Level 1A SPoTdata rather than the Level 18 data available for the project. Also, there was some uncertainty-as to whether the ioordinates of pass points could be established to an accu- racv of about j 1 pixel over the entire data set. For these rea- ro*, it was decid-edto rectify each of the SPoTimages individually, and then join them together to create the mo- saic (seeFigure 1 for SPOTscene locations). The eig-htSPoT panchromatic imagesof 10-m pixel reso- lution were rectified to the UTM coordinate system using six to 15 GCPSper image obtained from the GPSsurvey and/-or digitized from the existing uSGs1:24,000-scale topographic mips. All GCrs digitized from the uscs- maps were refer- enc-edto the North American Datum of 7927 (Nan zz)' These cCPswere converted to NAD 83 values using the U.S. Army Corps of Engineer's CoRPscoNprogram available from the tlcs offices in Rockville, Maryland' In a few instances,sup- plemental control in the form of small ponds or clumps of vegetation identifiable on the geocoded Landsat TM image we"reutilized to reinforce the rectification of the sPor images over the more remote areas of the Everglades National Park. An assessmentof the planimetric coordinate errors for each rectified sPor panchromatic image produced RMSE*,values ranging from i 5 to + 9 m, with the averagenvse for the -r7 +,0_.7 eig[t- sienes equivalent to m (i.e., pixel). The sPor image mosaic was created from the individual sPoT scenesusing the Desktop Mapping System (nvs)ru soft- ware package.naih digital srbt image (or tile)-was placed at its coirect Iocation (accordingto the UTM coordinatesof the upper left corner pixel) within a coordinaieAox for the en- tiie study area. An approximate 1-km wide by 60-km long. gap between two adjaient sPoT scenes (visible in Figure 1) i,vis patched with a strip of Landsat rv Band 3 image data (b) e"o"od"d and resampledto 1O-mresolution. Once the mo- Plale 2. Featuressuch as road intersections,tree is- iaic was assembled,a median filter was employed to mini- lands,and small pondscan be locatedon boththe clR mize banding and reduce artifacts along the margins of air photos(a) and the sPoTimage mosaic (b). The uTtvt overlap between adiacent images. In order to facilitate its groundcoordinates of such featuresare disitizedfrom use, a utM grid (Nan aa) with 5,000-mspacing was regis- ^ the spor imageand provideplanimetric control for the clR tered to the-mosaic (Figure 5)' The planimetric accuracy of aerial photographs. the geocoded mosaic was evaluated at 29 withheld control poinls (check points) and was better than + 1.5 pixels. The

PE&RS 1375 PEER.REVIEWED ARIICTE

l9w x,y digitizer coordinates for the annotated vegetation class boundaries to be transformed to uTM map coordinates (Easting, Northing) as the digitizing occurs. This procedure permits a segregationof tasks (interpretation, digitizing, edit- ing) and greatly "speeds-up" the development of the vegeta- tion database,Tests conducted to assessthe accuracyto which the features are digitized have yielded RMSE*.values of between + 5 and + 1Om. I I VegetationClassiflcation and Field Vedflcation with GPS Assisted Helicopter I Surueys A Ll I A hierarchical system of vegetationclassification was devel- l>=--+t I oped (and is still being refined) for the Parks that includes ,-l,\\1 \r \- --\ information on plant speciesassociations, land use, degreeof r \\\- \"\.. I human influence, and damage caused by Hurricane An-cirew. t..\ I \\ i -t Existing vegetation classification schemes such as the USGS \"\ Land-Use and Land-Cover Classification Svstem for Use with Remote Sensor Data (Anderson ol., 1,s76),the'U.S. Fish "t and Wildlife Service Cowardin System for Wetlands and Deepwatei Habitats of the United Sfofes (Cowardin ef a1., 1979),the Florida Departmentof TransportationClassifica- tion System, and the Florida Biological Diversity Project ' Classification System IFBDP, 1994) do not contain the level \ 0v of detail required for this vegetation mapping project. It was o"------1--P * \t --- - '\ thereforedecided saie 2< that a newclassification syltem should (1) contain classesat the individual speciesor speciesassocia- tion level that can be identified from the crR aerial photo- graphs and (2) have a hierarchical organization that allows Figure4. Distributionof 23 ccps established by a cps classesto be readily collapsed for compatibility with existing surveyin 1994 alongthe roadsthrough the Everglades vegetation classification schemesbeing used elsewhere in NationalPark and Big CypressNational Preserve. south Florida. The new EvergladesVegetation Classification System is being developedin conjunction with personnel from both the NPSand srwlrn and includes mo,saicoccupies slightly more than 200 Mbytes of disk space nine proposed classesof plant community and is resident on NPScomputers at Everglades National- types/land uses and an additional category Park.

Interprctationof Aerlal Photographs fw the Development of a Vegetation Database The timely development of an accurate,detailed vegetation databasein GISformat for the Parks requires the use of re- motely sensed data of sufficient resolution to identify and delineate vegetation classesto an accruacy of approximately 90 percent or better on one-hectareor larger plots, and of ad- equate geometric fidelity to lccate class boundaries and par- cels to accuracies generally compatible with Ntrlas for 1:24,000-scalemap sheets.These requirements precluded the use of Landsat TM (28.5 m) or SpoTHRV (20 m) multispectral imagery for the development of the vegetation database.In- stead, the use of the Napp and Nasa crR aerial photographs of 1-m (or better) resolution was consideredessential for the project. In order to expedite the interpretation of the photo- graphs and facilitate the construction of vegetation coverages in digital format, vegetation classesare delineated directly on cIR paper print enlargements (4x) produced from the CIR film transparencies. In addition, point features identified on the SPoTimages and the air photo enlargements are anno- tated and numbered, and their UTM coordinates(obtained Figure5. A portionof the spoTimage mosaic (10-m spa- from the geocodedspor images)are employed to establisha tial resolution)at the easternboundary of the Everglades cCP file for each photograph (Plate Z). NationalPark with 5,000-m urM grid (NAD83) superim- The annotated features on the photographs are then digi- posed. Notethe agriculturalland use adjacentto the tized, beginning with the GCPs.By digitizing the ccps first, a Park. set of photo rectification coefficients are generated which al-

1376 PEER.REVIETYED ARTICTE

Tlete 2. PRoposEoVEcEretrol Crnsses ron Sours FlontonNattol,qr Pnpxs In addition to the floristic characterizationof the natu- AND PRESERVES rally occurring Evergladesplant communities, the classifica- tion system includes categoriesof invasive exotic plants, A. Forest (onv) 1. Buttonwood Forest indicators of human influence such as off road vehicle 2. MangroveForest and air boat , and four hurricale damageclasses. In this 3. Pine Forest way, the Evergladesvegetation database will convey_details 4. Hardwood Forest on current vegetation,as well as human impacts and epi- B, Scrubland sodic disturbancesthat influence vegetationspecies distribu- 1. Buttonwood Scrub tions. 2. Mangrove Scrub The detailed classificationsystem requires that the inter- C. Hammock preters be familiar with the classification scheme and the Trooical Hardwood 1. Hammock as well as extremely skilled 2. Oak--SabalHammock iharacteristics of the vegetation, 3. Fan Palm Hammock in photo-interpretation.Attainment of a high level of classifi- 4. Buttonwood Hammock cation accuracyalso requires the interpretersto have direct 5. Sabal Palmetto Hammock associationwith the plants and to make cross-correlations D. Savannah between the photographic imagesof the vegetationclasses 1. Palm Savannah and their appearancein the . In this con- 2. CypressSavannah text, fieldwork is greatly facilitated by the use of NPSBelI Jet 3. Pine Savannah Ranger 206 helicopters that are available for ground truth E. Swamp collection and verification of image interpretations.However, 1. Mixed Hardwood Swamp ("- and a pro- 2. CypressSwamp helicopter flight time is expensive $500.00/hr), 3. Bav Head Swarno cedure involving use of the sPor image mosaic and the latest q. Willow Head Swamp technology in laptop computers,GPS receivers, and image F, Prairie/lVlarsh processing/positioning/displaysoftware has been developed 1. Wet Prairie io expedite data collection and verification. The helicopters 2. Dry Prairie are equipped with cPSreceivers that enable the pilots to pre- 3. Halophytic HerbaceousPrairie define their flight track, to conduct real time navigation 4. Broad LeavedEmergent Marsh guided by the GPSunit, and to record the coordinatesof G. Exotics landing points or featuresof interest.In order to maximize (caieput) 7. technology,the srot image 2. Casuarinaspp. (Australian pine) the advantageof this positioning 3. Colubrina asiatica (lather leafl mosaic of LO-mpixel resolution was aggregatedto a mosaic 4. Schinusterebinthifolius (Brazilian pepper) of 2o-m pixel resolution and divided into four tiles, each H. Open Water about 13 Mbytes in size. These image files were then loaded 1. Lakes/Ponds into a Dell Latitude xP 486 laptop computer (1oo Mhz) along 2. Rivers/Streams with the DMS(R-WEL, Inc,) and Field Notes (Pen Metrics, 3. Canals Inc.) softwarepackages. A Trimble Pathfinder Professional 4. Air Boat Trails (six-channel)cPS receiver with an external antennamounted L Non-Vegetated on the forward hull of the helicopter was then connectedto 1. Beaches This set-up enablesa 2. Mud (Tidal) Flats the serial port of the laptop computer. 3. ExposedRock (i.e., pinnacle rock) person in the rear seat of the helicopter to hold the computer 4. Off Road Vehicle (ORV)Trails on his or her lap, display the satellite image mosaic, and (Plate f. Special Modifiers: track in real time the flight path of the helicopter 3)' 1. Density Classes Most importantly, it provides a means of collecting ground 2. Hurricane DamageClasses truth information that is linked to coordinatesprovided by 3. Human Influence the GPSreceivet. Upon reaching an irea of interest,the heli- coDtercircles at low altitude and/or lands to allow identifica- tion of plants (Plate4). Speciesattribute information and of special modifiers (Table 2). A higher level of detail within additional notes pertaining to fire history or exotic control the plant community types includes plant associationsde- measures that may have influenced the Elreaare entered into fined by typical dominant species.For example, under Forest the computer and linked with the GPScoordinates. This pro- (type) there are four associations:Buttonwood, Mangrove, cedure also can be used with vehicle or foot surveys. Pine, and Hardwood. These associationsmay be further sub- The data gatheredduring the ground and helicopter sur- divided to include classesfor individual species,for exam- veys Erreused to verify vegetationinterpretations from the ple, (red mangrove), Avicenni.a getmi- 1,994/1,9s5uscs NAPPair photos. After verification, the nans (black mangrove),Laguncularia racemosa(white digital vegetationboundary files, along with transportation malgrove), and Mixed Malgrove within the Mangrove Forest and hydrographic data in digital format, are input to the ARC/ association.The speciesdescriptions for the EvergladesCIas- INFOsoftware package resident on an IBMRISC System 6000 sification Systemwere compiled from Craighead(1971), Mc- workstation, edited, attributed, and edge matched to create Pherson (1973),Davis and Ogden (1994),and severalNPS the cIS database.Tiles correspondingto the USGS1:24,OOO' South Florida ResearchCenter Reports (1980-1986)for the scale topographic quadrangleseries are then plotted as hard- Evergladesand Big CypressNational Parks/Preserves(e.9., copy maps. Gundersonand Loope,1982; Olmstead et al., 1983).The classificationinformation provided here is preliminary and is VegetationMaps and Ptojected Database Use used to help clarify the photo-interpretationand mapping The crs databaseand vegetationmaps are currently being proceoules. constructedwith an anticipated completion date sometime in

PE&RS PEER.REVIEWED ARIICTE

digital data sets to study the relationships between human activities, hurricane damage,and vegetationpatterns. This information, in turn, will provide a basis forlssessing future changesdueJo the increaJingpressures facing the paiks. tn addition, higher_resolutionimage subsetsof important com- munity and landscapetypes and featureswill be established to supp-ort a long-term environmental monitoring program for the Parks to assesschanges at the individual speiies ind community levels. These image subsetswill allow the parks to monitor the relationship of vegetationchange to other en- vironmental variables. Conclusion The development of a digital databasein GrSformat for the Parks of south Florida is made possibleby the availability of recent SpoTpanchromatic satellite image data of 10-m reso- Iution, cIR aerial photographsrecorded-in 1994-1995,and new approachp,sto the integration of cps, image processing, and cts technologiesto facilitate rectification,inilysis, and verification of data. It is envisioned that the geocodedsror image mosaic will provide NpSpersonnel with a quick and reliable means of plnpointing f"itnrur of interest ind devel- oping image subsets that will aid Park managers,scientists, rangers,and law enforcementofficers in their various roles. The cIS databaseand associated1:24,000-scale hardcopv vegetationmaps will provide the NpSwith the first compre-- hensive, up-to-datespatial data to support park management personnel in evaluating the status of vegetationand the threats causedby urban expansion,the intense use of border- ing agricultural lands, and the encroachment of exotic plant species.In particular, the databasewill provide a means for assessingdamage caused by natural stochasticevents such as Hurricane Andrew. The methodologiesemployed in this pro- ject, which include the construction of a geocodedsatelliie

Plate3. Collectionof groundtruth is facilitatedby the use of helicoptersand by employinga laptopcomputer interfacedto a cps unit and softwarefor imagedisplay/ attributeentry. The cps continuallytracks the helicopter. Thistrack is displayedon the spoTimage mosaic and at- tribute informationis recordedin real time.

1996. As the project proceeds,the digital and hardcopy products are being transferredto the South Florida Niiural ResourceCenter, Everslades National Park. and to the GIS units at the Headquartersof Big CypressNational Preserve and BiscayneNational Park. Two preliminary samplesof vegetationmaps are presentedin Plate b. The first depicts Pine Savannahvegetation in the Long Pine Key area of Ever- gladesNational Park that suffered stem breakageand defolia- tion by the strong winds and tornadoesassociated with Hurricane Andrew (PlastebA). Although considerabledam- age was noted on the 1992 NASAphotographs, recovery of the pine trees was observedduring helicopter surveys con- ducted in 1995. The impact of human actfuities on vegeta- Plate4. The Universityof Georgiaand rupSpersonnet ex- tion patterns,on the other hand, may be more long lasting. amine SawgrassWet Prairievegetation in Big Cypress For example, evidence of past row crop planting of vegetable NationalPreserue. The use of helicoptersand real-time farms abandonedmore than 40 years ago remains in Si- position/dataentry greatly expedites field verificationand parks wgrassWet Prairie and CypressForest areas in the providesa computerrecord that is used in the laboratory (Plate 5B). to facilitatephoto interpretationof vegetationclasses. Park researchersand managerswill use the maps and

1378 PEER.REVIEWED ANTICTE

SouthFlorida Park VegetationDatabase

VegetationClasses

I Pin" Forest/CypressForest Pine ForestDamage Class I >0 to 25?o damaged E )25 to 50?o damaged E >50 to 75% damaged n )75 to 100o/odamaged ffi Ui*"d Wet Prairie-Abandoned Agriculture & HardwoodScrub I rropical HardwoodHammak ffi rinulcypreseSavannah ffil r*otio (Schinus) ffi sayhead/WillowheadSwamp I ruunbnbergia/CladiumWet Prairie ffil Hurn"n Influerre ffi PrincipalRoads

N L@tor lvtap 1

Sale B Plate5. Preliminaryvegetation maps for portionsof EvergladesNational Park and Big CypressNational Pre- serve.(A) PineSavannah vegetation in a portionof EvergladesNational Park classified according to levelsof damagecaused by HurricaneAndrew. (B) Evidenceof humaninfluence and abandonedagricultural fields de- tected in Big CypressNational Preserve communities of MixedWet Prairie.

PE&RS PEER.REVIEWED ARTICTE

image mosaic to use in coniunction with aerial photographs, A historical perspective, Everglades: The Ecosystem and Its Res- and the combination of cps, helicopter, and ground survey torution (S.M. Davis and J.C.Ogden, editors), St. Lucie Press, techniquesto verify photo interpretation,can be used in Delray Beach,Florida, pp. 47-84. other remote and/or inaccessibleareas to facilitate the gener- Loope, L., M. Duever,A. Herndon, J. Snyder, and D. Jansen,1994. ation of detailed databasesvital for the preservationof Hurricane impact on uplands and freshwater swamp forest, Bio- unique environments. scienc e, 44(4):238-246. McPherson,8.F., 1973. VegetationMap of Southern Partsof Sub- Acknowledgments areas A and C, Big CypressSwamp, FLorida, Hydrologic Investi- gationsAtlas HA-492, This researchproject is sponsoredby the U,S. Departmentof Florida Departmentof Natural Resources and U.S. GeologicalSuruey, Washington, D.C., 1 p. Interior, (Nns),through Cooperative Olmstead,LC., W.B. Robertson, and O.L. Bass, Agreement#5280-4-9006. The assistanceof the Florida De- '1.983. fr., f. |ohnson, Jr., Vegetation of Long Pine partment of Environmental Protection, Petersburg, Key, Everglades National Park, St. Flor- Technical Report SFRC-83/05,South Florida ResearchCenter, ida; the South Florida Water ManagementDistrict, West National Park Service,U.S. Departmentof Interior. Homestead. Palm Beach,Florida; SPOT Image Corporation,Reston, Vir- Florida,64 p. ginia; and Survey ResourcesInternational, Inc., Houston, Pimm, S.L. G.E.Davis, L. Loope, C.T. Roman,T.J. Smith III, and J.T. Texas is gratefully appreciated.Mr. William Hanis of Global Tilmant, 1994. Hurricane Andrew, Bio science, 44(4')1224-225. Satellite Surveys,Decatur, Alabama, worked tirelessly to en- Pumpelly, J.W., 1967.Color-separation and printing techniquesfor sure that an accuratecPS survey was accomplishedon photomaps,Suweying and Mapping, 27(2):227 -28O. schedule. Rose,M., and F. Draughn, 1991.GIS applications at EvergladesNa- tional Park, GIS World, a(3):a9-51. Refelences Rutchey,K., and L. Vilchek, 1994.Development of an Everglades vegetation map using a SPOT image and the global positioning Anderson, J.R.,E.E. Hardy, l.T. Roach,and R.E. Witmer, 1976.A system, Photogrammetric Engineering & Remote Sensing, 6O(6): Land Use and Land Cover Classification Svstem for Use with Re- 767-775. mote SensorDota, U.S. GeologicalSurvey, Professional Paper Spradley,L.H., 1994.Cost-effective data for regional GIS bases,Pro- 964, U.S. GovernmentPrinting Office,Washington, D.C., 28 p. ceedings of the ISPRSCommission IV Symposium on Mapping Cowardin, L.M., V. Carter,F.C. Golet, and E.T. LaRoe,7979. Classifi- and Geographic Information Systems (R. Welch and M. Remil- cation of Wetlands and Deepwater Habitats of the llnited Stufes, lard, editors), International Archives of Photogrammetry and Re- Fish and Wildlife Service,U.S. Departmentof the Interior, mote Sensing,Athens, Georgia, 30(Part 4):2O3-2O8. Washington,D.C., p. 103 Welch, R., 1985.Cartographic potential of SPOT image data,Phofo- Craighead,F.C., Sr., 7971. The Treesof South Florida, Vol. L The grammetilc Enginee ring & RemoteS e nsing, 51 (8) : 1085-1 091. Natural Environments and Their Succession,Universitv of Mi- Welch, R., T.R. |ordan, and M. Ehlers, 1985.Comparative evalua- Press, ami Gables,Florida, 212 p. tions of the geodeticaccuracy and cartographicpotential of Davis, S.M., and J.C.Ogden (editors),1994. Everglodes: The Ecosys- Landsat-4/-S TM data, Photogrammetric Engineering & nemote tem and its Restoration,St. Lucie Press,Delray Beach,Florida, Se ns ing, 57(s)124s-Iz6z. 826p. Welch, R., M. Remillard, and J. Alberts, 1992.Integration of GPS,re- Doren, R.F.,L.D. Whiteaker,G. Molnar, and D. Sylvia, 1990.Restora- mote sensing and GIS techniques for coastal resource manage- tion of former wetlands within the Hole-in-the-Donutin the Ev- ment, Photogrammetric Engineering & Remote Sensing, 58(11): erglades National Park, Proceedings of the Seventh AnnuaL 1571-1.578. Conferenceon Wetlands Restoration and Creation, Plant City, Welch, R., and M. Remillard, 1994.Integration of GPS,digital image Florida.pp. 33-50. processingand GIS for resourcemapping applications,Proceed- Duever,M.]., ].E. Carlson,].F. Meeder,L.C. Duever,L.H. Gunderson, ings of the ISPRSCommission lV Symposium on Mapping and L.A. Riopelle, T.R. Alexander,R.L. Myers, and D.P. Spangler, Geographic Information Systems(R. Welch and M. Remillard, 1986. Tfte Big Cypress National Preserve,Research Report Num- editors), International Archives of Photogrammetry and Remote ber 8, National Society,, New York, 455 p. Sensing,Athens, Georgia,3O(Part 4):7O-74. Florida Biological Diversity Project, '1994.Florida Biological Diver- sity Project (FBDP) Newsletter, Fall Issue, Florida Cooperative Fish and Wildlife ResearchUnit, University of Florida, Gaines- Roy Welch . ville, Florida, B p. Roy Welch received the B.S. degree in geography and biol- Gunderson,L.H., D.P. Brannon, and G. Irish, 1986. VegetationCover ogy from Carroll College (Wisconsin), the M.A. degree in Types of , Everglades National Park, Derived physical geography from the University of , and from Landsat Thematic Mapper Dafo, Technical Report SFRC- the Ph.D. degree in remote sensing, photogrammetry, and 86/03, South Florida ResearchCenter, National Park Service, cartography from the University of Glasgow, Scotland. He is U.S. Departmentof Interior, Homestead,Florida, 6 p. currently a Research Professor of Geography and the Director Gunderson,L.H., and L.L. Loope, 1982.An Inventory of the Plant of the Center for Remote Sensing and Mapping Science Communities within the Deep Lake Strand Area, Big Cypress (CRMS)at The University of Georgia. Dr. Welch is also Presi- National Preserve,Technical Report T-666, South Florida Re- dent of Commission IY, Mapping and Geographic Informa- searchCenter, National Park Seivice,U.S. Departmentof Inte- tion Systems, International Society for Photogrammehy rior, Homestead,Florida. 39 p. and Remote Sensing fispns). He is a past-President of the Ameri- fensen, K. Rutchey,M.S. Koch, and S. Narumalani, 1995.Inland f.R., can Society for Photogrammetry and Remote Sensing wetland changedetection in the EvergladesWater Conservation (ASPRS). Area 2.A.Using a Time Seriesof Normalized RemotelySensed Dala, Photogrammetfic Engineering & Remote Sensing, 61.(2): 199-209. Marguerite Remillard Lenzen,T.W., and T.W. Foresman,1993. Digital image databases Marguerite Remillard received the B.A. and M.A. degrees in support GIS operations,GIS World, 6(11):36-38. biology from the State University of New York, and the Ph.D. Light, S.S.,and l.W. Dineen, 1994.Water control in the Everglades: degree in ecology from The University of Georgia. She is cur-

1380 PEER'REVIEWED ARIICTE

rently an Associate ResearchScientist at the Center for Re- Robert Doren mote Sensingand Mapping Science(cnvs), The University Robert Doren received the B.S. and M.S. degreesin botany of Georgia.Dr. Remillard also servesas Secretary,Commis- from Florida State University and the University of Mary- sion IV, Mapping and Geographic Information Sysfems,In- land, respectively.He is currently the AssistantDirector of ternational Society for Photogrammetry and Remote Sensing the South Florida Natural ResourceCenter, Everglades Na- (rspns). tional Park. Mr. Doren is responsiblefor the operation of the Center,and for researchprojects with outside agencies.

F(l RIH C()M I IIG ARTI CI.ES

The followinglist includesonly thosealticles scheduled for Tang Uang and C/rnstlonHeipke, Automatic Relative publication in PEhRS through April 1996. Orientation of Aerial Images. Donald L. Light, Film Camerasor Digital Sensors? The R. M. Batsonand E M. Eliason,Digital Maps of Mars. ChallengeAhead for Aerial Imaging. Michael F. Baumgaftner and Albert Rango, A Donald C. Rundquist,Luoheng Han, lohn F. Schalles, and Microcomputer-BasedAlpine Snow Cover Analysis Ieffrey S. Peake,Rernote Measurement of Algal Systern(ASCAS). Chlorophyllin SurfaceWaters: The Casefor the First Michel Boulianne, Rock Santene, Paul-Andft Gagnon, and Derivative of Reflectancenear 690 NM. Cl6ment Nolette,Floating Linss aud Conesfor L-Iseas SorcnRyherd and Curfis Woodcock,Combining Spectral a GPSMission Planning Aid. and Texture Data in the Segmentationof Remotely Gregory Catbone, Sunil Narumalani, and Michelle Kng, l. SensedImages. Applicationof RemoteSensing and GIS Technologies Eric M. Sorrderr,Carlton M. Britton, and H. Everitt, with PhysiologicalCrop Models. /omes Total Ground-CoverEstimates from CorrectedScene Allen E. Cook and John E. PinderIII, Relative Accuracy of RectificationsUsing CoordinatesDetennined frorn BrightnessMeasurements. Maps and GlobalPositioning Systems. Howatd Schultz,Shape Reconstruction flom Multiple Christopher Deckert and Paul V. Bolstad, Forest Canopy, Imagesof the Ocean Surface. Terrain, and DistanceEffects on Global Positioning Tian-YuanS/ulr, A PhotogrammetricSimulator for Close- SystemPoint Accuracy. RangeApplicatious. Bon A. Dewitt, Initial Approxinrationsfol the Thlee- Mohamned E. Sholo',Inurence l. Wilson, and Dwayne L. Dimensional Conformal Cooldinate Transfonnation. Surdu-Miller,Effect of Radar Parameterson Sea Ice Sam Eksfrund,Landsat TM-BasedForest Dantage Tonal and Texttrral SignaturesUsing Multi-Frequency Assessmeut:Colrection for TopographicEffects. Polarimetric SAR Data. YasserEl-Manadili and Kurf Novok, PrecisionRectificatiou Inwrcnce V. Stanislawski,Bon A. Dewitt, and Romesfi L. of SPOT ImageryUsing the Direct Linear Shrestrla,Estirnating Positional Accuracy of Data TransfonnatiouModel. Layers within a GIS through Er:rorPropagation. Maurice S. Gyer,Methods for Computing Photograrnmetric SteveStehman, Estimating the Kappa Coefficientand Its RefractionCorrections for Vettical and Oblique Valiance Under Stratified Random Sampling. Photographs. R. Hutchinson, and P. Christian Heipke, Wolfgang Konrus, and Arifon Daniel Steinwand, John A. John Pfannenstein,The Evaluation of MEOSS Airborne Snyder,Map Prolectionsfor Global and Continental Three-LineScanner Lnagery: PlocessingChain and Data Setsand au Analysis of Pixel Distortion Caused Results. by Reprojection. Matthew Heric, Cauoll Lucos, and Christopher Devine, Tlte futtonio Maria Garcia Tommaselli and Cl6sio Luis Tozzi, A Open SkiesTleaty: QualitativeUtility Evaltrationsof RecursiveApproach to SpaceResection Using Aircraft Reconnaissanceand CouunercialSatellite StraightLines. Imagery. Paul C. Van Deusen,Unbiased Estimates of Class Peter E. loia and lanet C. /orgenson, Conrparison of Three Proportionsfrom ThernaticMaps. Methods for Mapping Tundla with LandsatDigital Howard Vercgin, Error Propagation through the Buffer Data. Operation for Probability Surfaces. RonaldKryok and Tom Baltzer,The GeophysicalPlocessor Gary M. I,Vohl,Operational Sea Ice Classificationfrom Systemat the AlaskaSAR Facility. Synthetic Apertule Radar Imagery. RongxingLt, Design and Implementation of a Ding Yuan, Natural Colrstraintsfor InverseArea Estimate PhotograrnnretlicGeo-CalcuIator a in Windows Couections. Environment.

PE&RS