GAP Bulletin Number 5
June 1996

State Reports

(Note: You need to wait for this page to completely load before choosing a state.)

New Hampshire
New Jersey
New Mexico
New York
North Carolina
Rhode Island
West Virginia


All primary GAP data layers were completed in spring 1995. New project teams at the University of Arizona (UA), Tucson, and Northern Arizona University (NAU), Flagstaff, have been funded to conduct accuracy assessment and analysis and to develop a final report. In addition, Arizona is part of two 4-state ecoregion projects. The Four Corners project includes parts of New Mexico, Colorado, and Utah comprising the Colorado Plateau and Southwest Highlands ecoregions. This project will be coordinated by NAU. The Mojave ecoregion project includes California, Nevada, and Utah and is well under way. Assessment of the vegetation map is partially funded by the State of Arizona Department of Game and Fish, which is an active cooperator in the project. The data set will be shared with the state and others through the NBS National Biological Information Infrastructure (NBII) with funding provided by the NBS Division of Information and Technology Services.

Bill Halvorson
University of Arizona, Cooperative Park Studies Unit, Tucson, (520) 670-6885

Kathryn Thomas


The Arkansas GAP project has completed its statewide land cover, land ownership, and management status maps. The land cover map has been aggregated, and accuracy assessment is now under way. The land cover data have been aggregated using the "Montana" method, which has worked well, given the spectral classification procedures used initially. Data have been aggregated at 2, 10, 40 and 100 hectare levels, with the 100 ha level to be provided to the National GAP effort.

Increased discrimination for the urban mapped areas has been obtained by integrating GAP spectral data with TIGER and STF (Standard Tape Format) data to isolate urban classes from bare rock and similar spectral signatures that are not urban. Cartographically appealing hardcopy products have been developed in addition to the digital maps. Land cover classes have been assigned to the spectral classes using a variety of data sources, including SOFIA, USFS, and state agency data. Twenty percent of the Game and Fish plot data (3,000 plots) and a similar percentage of the U.S. Forest Service's stand/compartment data were excluded for consideration in the information class assignment phase. These plots and compartments/stands were reserved on a random selection basis and were used later in the accuracy assessment phase.

In addition, an important accuracy assessment data set was developed by selecting 2,000 random field plot locations and providing them to the Arkansas Forestry Commission. The Arkansas GAP team developed an innovative method for site selection based on the Utah project and a software application that produces a small hardcopy map for each quad. These maps were given to Forest Commission field staff along with disposable cameras. The field staff recorded the forest types based on the Arkansas GAP land cover classification and made their own comments along with sample photographs for a permanent record.

Land ownership and management status mapping was also conducted in close cooperation with participating agencies. Less than 10% of the state is under public management or owned by natural heritage groups. For the state, less than one-half of a percent is in GAP management class one, less than two percent is in class two, seven in class three, and the remainder in class four (or is unknown). As the land ownership/management effort went forward, it became clear that large portions of the state are managed for various natural and/or wildlife purposes, such as hunting clubs, but are in private ownership. Unfortunately, the location and character of these private lands are unknown and could not be included in the current GAP effort.

Vertebrate distribution maps have been developed for birds, and the work for mammals and herps is largely done. Its final completion has been delayed due to the serious illness of a key participant in the statewide steering committee. Vertebrate distribution accuracy assessment will begin upon completion of the mapping effort. The project focus is now shifting to the development of a detailed final report, metadata, and procedures for the distribution and maintenance of the data. A CD-ROM for distribution is planned along with reports and maps. Availability of the data on the Web is also planned.

W. Fredrick Limp
Center for Advanced Spatial Technologies
University of Arkansas, Fayetteville, (501) 575-6159


Project Status

The California Gap Analysis is nearing completion. Land cover and land management mapping have been virtually completed for all ten planning regions of the state, and we are shifting attention towards data analysis and distribution. In 1996, we will be incorporating revisions of the land cover data suggested by reviewers and then publishing a final report. The current plan is to distribute the database and regional reports on both the World Wide Web and on a CD-ROM. These two media would be used both for locating and transferring data and for common queries of the database. Discussions are still under way to determine who within the state should be responsible for long-term maintenance and distribution of GAP data. The Universal Resource Locator addresses (URL) for California GAP data are: (for information about GAP and other related research projects at UCSB) and (for accessing GAP data as they are completed).

In 1995, one Ph.D. dissertation was completed (Pete Stine) and another is in draft (Kathryn Thomas). Besides two papers at the GAP symposium in Charlotte, NC, last year, several peer-reviewed articles relating to GAP were published or accepted for publication since the last newsletter:

Church, R.L., D.M. Stoms, and F.W. Davis. 1995. Reserve selection as a maximal covering location problem. Biological Conservation, in press.

Davis, F.W., P.A. Stine, and D.M. Stoms. 1994. Distribution and conservation status of coastal sage scrub in southwestern California. Journal of Vegetation Science 5: 743-756.

Davis, F.W., P.A. Stine, D.M. Stoms, M.I. Borchert, and A.D. Hollander. 1995. Gap analysis of the actual vegetation of California: 1. The Southwestern Region. Madroņo 42: 40-78.

Hollander, A.D., F.W. Davis, and D.M. Stoms. 1994. Hierarchical representation of species distributions using maps, images, and sighting data, in Mapping the Diversity of Nature, R. I. Miller, editor, Chapman & Hall, pp. 71-88.

Sierra Nevada Ecosystem Project

The California Gap Analysis (CA-GAP) project participated in the Sierra Nevada Ecosystem Project (SNEP) funded by the U. S. Forest Service. The mission of SNEP was to (a) define the spatial extent and dynamics of key structural, functional, and compositional features of the ecosystem; (b) identify the benefits humans draw from it; and (c) identify management alternatives and their effects on ecosystem integrity and its sustained capacity to provide the full range of benefits. A Gap Analysis of plant communities was one of several analyses of regional biodiversity. Others included late-successional/old growth forests, aquatic species and habitats, and significant ecological areas, although these were generally limited to public lands because of data availability. Because of the vast amount of spatial data compiled for SNEP, we were able to refine the standard GAP management class definitions to include permitted land uses on public lands such as grazing allotments and commercial timber harvest from forest plans. The final SNEP report to Congress, including our chapter on GAP, is in press.

Biodiversity Management Area Selection

Besides Pete Stine's dissertation, the Biogeography Lab at UCSB undertook two additional studies of reserve selection algorithms. The first study reformulated the reserve selection models in the conservation literature as a classical Maximal Covering Location Problem (MCLP) as described in the operations research and regional science literature twenty years ago. The MCLP can be solved optimally (that is, no better solution exists), and most problems of the size described for reserve selection can be solved with reasonable computer resources. We solved the MCLP model for a real application using vertebrate distribution data prepared for the Gap Analysis of southwestern California. The areas are defined by each of the 280 7.5' quadrangles, and the regional species pool contains 333 native vertebrates. Therefore, the species-site matrix consists of 280 columns by 333 rows. A maximum of twelve sites are required to cover all species at least once, although 327 species can be covered in just 7 sites. "Solutions" on an IBM RS-6000 workstation took an average of 2.8 seconds of CPU time over the 12 solutions, with none taking more than 9 seconds. This analysis is being published in Biological Conservation (see Church et al., above).

The second study recognized the limitations of the simple "covering" model. An optimization model was developed for SNEP (in collaboration with Rick Church and B. J. Okin at UCSB and Norm Johnson at Oregon State) that selects new areas for biodiversity management. The model, dubbed BMAS (Biodiversity Management Area Selection), minimizes total area and maximizes overall suitability of the selected sites while meeting predefined levels of representation (e.g., 10%) for each community type (or wildlife species). Suitability was defined for this study by habitat quality (road density and human population density) and management factors (proportion of private ownership and the degree of fragmentation of land ownership). These factors were chosen for their potential impact on biodiversity management as well as the availability of data across the entire ecoregion. The SNEP study was not authorized to make formal management recommendations. Instead, we used BMAS to explore the dimensions of the problem by looking at a number of alternatives. The alternatives varied by their assumptions (which lands are considered currently protected), target levels of representation, biodiversity elements (communities versus vertebrate species), and the suitability factors. This study will be included in the SNEP report to Congress, and two journal articles are in preparation.

David Stoms
University of California, Santa Barbara, (805) 893-7655


The final draft of the land cover layer for Colorado has been completed and will be under review through this summer. Current efforts are dealing with edge-matching with Utah, Wyoming, and New Mexico and with development of the data dictionary as part of a manual explaining the processes followed. The land cover layer will be further reviewed by field biologists and vertebrate modelers in the subsequent months for revision during the final six months of the contract period. Accuracy assessment is planned to begin this summer in coordination with Wyoming and will use aerial video.

Meanwhile, Colorado Gap Analysis vertebrate modeling efforts began in earnest with assemblage of the modeling team. A team of six biologists and ecologists has joined Colorado Division of Wildlife (CDW) staff, and we are now building links between data sets and coverages available through the CDW's Wildlife Resource Information System (WRIS) and the land ownership and vegetation data coverages developed specifically for Colorado's Gap Analysis Project.

The main data fields being linked from the Colorado Division of Wildlife's species database for GAP modeling efforts relate to county-based distributional records, wildlife habitat relationship information, and physical habitat descriptors related to species' environmental requirements and life histories. Ancillary information from other WRIS databases will provide more site-specific information from the Scientific Collections Permit Database (SCICOLL), the Herptiles Observations Database, the Colorado Raptor Database (CORAPTOR), the Aquatic Database Management System (ADAMAS), and the Colorado Latilong Distribution Studies (WILDATA).

Additional information from partnering groups will come from the Colorado Bird Atlas Project (CBAP) and the Colorado Wildlife Heritage Database (with the Colorado Natural Heritage Program). The Natural Resources Ecology Laboratory at Colorado State University will provide integration opportunities with the Division's Species Ranking Project and provide additional information for GAP project use in evaluating management considerations relative to protection of Colorado's biodiversity. Special thanks go to the state programs in Wyoming and Utah for sharing observations from their Gap Analysis endeavors, and New Mexico for evaluation of the commonality of land cover mapping efforts in New Mexico, Colorado, Arizona, Utah, and Wyoming.

Donald L. Schrupp
Colorado Division of Wildlife, Denver, (303) 291-7277

Dr. William Reiners and Tom Thompson
University of Wyoming, Laramie, (307) 766-2235


(see Massachusetts, Connecticut, and Rhode Island)


(see Maryland, Delaware, and New Jersey)


The Florida Gap Analysis Program has completed land cover classification of about a fourth of the state, from Lake Okeechobee through the Keys. This area is now being reviewed for a final iteration of the classification. The TNC Southeastern Region classification scheme was used to delineate classes to the alliance level or better. Analog videography played a major role in providing the volume of ground data necessary for a detailed classification at the alliance level. We expect to complete about half the state (all lands south of Orlando) by early summer of 1996. This effort is being greatly assisted by additional funding from the Florida Game and Fresh Water Fish Commission. Ground-truthing and mapping assistance have also come from a variety of agencies and individuals, most notably the U.S. Army Corps of Engineers. The NOAA C-CAP National Ocean Program is gearing up to also offer significant assistance in a cooperative effort to map Florida's land cover and land cover change. Auxiliary mapped information (existing land use, NWI, and SCS county-level soils) is being prepared ahead of classification. Coverages for over half the state have been reviewed, cleaned, and/or modified.

We are mapping/modeling all native and exotic terrestrial vertebrate species in Florida, as well as butterfly, skipper, and ant species. Breeding and wintering birds in the state are being treated separately. Distribution maps for all species are now being externally reviewed for accuracy. Distributions were determined from museum and other records and published literature, with interpolation and extrapolation used as necessary. Databases of species' habitat use are complete for all species except birds, which are nearly complete. Habitat information was compiled from over 1000 sources. Our next step will be to generate a matrix of species with habitat, utilizing the TNC classification scheme. Information necessary for species-specific modeling, where available, has been collected for most species. This information includes home range size and dispersal distances.

Leonard Pearlstine
FL Cooperative Fish and Wildlife Research Unit
University of Florida, Gainesville, (352) 846-0630


The State of Hawaii Department of Land and Natural Resources (DLNR) was awarded a National Biological Service (NBS) State Partnership Grant to utilize satellite imagery for land cover classification and mapping. A major objective of the project is to lay the groundwork for the re-establishment of GAP in Hawaii. Satellite imagery covering the major islands of the Hawaiian archipelago is being provided by SPOT. The DLNR is one participant in a group-purchasing program which includes 26 state, federal, and private participants.

Due to problems with cloud cover over significant portions of the islands, SPOT has not been able to provide coverage in the time frame originally agreed upon. We are hopeful that complete coverage will be available by February 1996. Currently, project participants are setting the groundwork for the project, which will begin in earnest once the imagery is available.

Ronald Cannarella
Department of Land and Natural Resources, Honolulu, (808) 587-0166


The Idaho Gap Analysis group is continuing to analyze the original 200 ha MMU data set while remapping the state using a MMU of 2 ha. This land cover remapping is a cooperative effort of the U.S. Forest Service and others. Roly Redmond's GAP group in Montana and Tom Edwards's GAP group in Utah are doing the actual mapping and completed the draft map on March 15. The GAP group at Idaho is revisiting the vertebrate models using ecological themes unavailable in 1989. Work is being conducted in conjunction with the Idaho Heritage Program. We have plans to compare maps of predicted vertebrate ranges at different levels of spatial and thematic detail.

Gerry Wright and several co-authors have published articles on the use of the Idaho GAP maps for assessing what additional protection of cover types and vertebrate species is obtained under various proposals for new National Parks and Wilderness areas (Merrill et al. 1995). The Idaho vegetation map has been edge-matched with the Oregon and Washington maps by Michael Murray, Troy Merrill, Kelly Cassidy, and Blair Csuti. They were able to map 95 cover types in three states, mostly at the undifferentiated level. Chuck Peterson and graduate students at Idaho State University conducted a two-year field check for herptiles on Craig Mountain as part of a master's thesis. Overall accuracy was 85%, compared to 65% for the best available field guides. The results are being written up for publication. Now that we have a 2 ha land cover map, we will compare predictions for vertebrate species at the two very different MMUs.

Troy Merrill, Mike Murray, and Mike Scott are examining the distribution of cover types within special management areas to assess representation across the entire geographical and ecological range of each cover type. Mike Jennings, Patrick Crist, and Angie Evenden (US Forest Service Natural Areas program manager for eight western states) will conduct a conservation assessment of US Forest Service Research Natural Areas by comparing their distributions and sizes with GAP data. Kevin Gergely is just starting a project that asks which biological and ecological processes can be accommodated on differently sized management areas and, if areas are too small, what ex situ or transboundary management activities are required to maintain biological objectives of the area.

Dave Mattson and Troy Merrill have used the GAP land cover maps and maps of human activity to predict habitat suitability in areas of potential conflict between bears and humans. These maps are being used to help design conservation strategies for grizzly bears in Idaho.

As part of the second generation GAP effort in Idaho, we are collaborating with the Idaho Department of Fish and Game, Heritage Program, and The Nature Conservancy to revise the vertebrate distribution maps using hexagons as our unit of geographical occurrence. Plans are under way for a collaborative effort with adjacent states and The Nature Conservancy to produce a monograph of the ecological and cultural features of Bailey's ecoregions.

Literature Cited

Merrill, T., R.G. Wright, and J.M. Scott. 1995. Using ecological criteria to evaluate wilderness planning options in Idaho. Environmental Management 19:815-825.

J. Michael Scott
ID Cooperative Fish and Wildlife Research Unit
University of Idaho, Moscow, (208) 885-6336


The Illinois GAP Project started in November 1995. The Illinois Natural History Survey (ILNHS) received Thematic Mapper (TM) scenes for the state to conduct the Critical Trends Analysis Project (CTAP). The "first cut" vegetation map for Illinois was produced for CTAP using a Boolean mask and vector field segementation.

Landscape Stratification

When utilizing full TM scene data, experience has shown that clustering algorithms often fail to adequately discriminate landscape elements which individually constitute small proportions of the entire scene. This is especially true where urban and built-up lands constitute a small overall percentage of the landscape. To ensure that the spectral signatures for the urban and built-up lands are properly characterized during the unsupervised training stage, these lands are extracted from each TM scene utilizing a Boolean mask created from the block-level, rural-urban classification contained within the 1990 Census TIGER/Line and STF1B files. This Boolean mask was subsequently used to perform two separate classifications, one for the urban portion and a second for the predominantly rural portion of the TM scene. Luman and Ji (1993) determined that a similar approach is effective in improving classification accuracy.

Image Segmentation

Conventional approaches to unsupervised image classification use pixel classifiers that assign unknown pixels to a spectral class using no contextual information. Thus the spatial domain, expressed within the image as geometrically homogeneous areas (e.g., agricultural lands), is ignored. Yet, such information is important to the conventional photointerpretative process. Research has shown that the inclusion of spatial structure into the classification process can improve discrimination when used for some remote sensing applications (Woodcock 1992, Nichol 1990). This approach used massively parallel deterministic relaxation optimization algorithms to partition an image into a set of regions which correspond to objects on the landscape, and is generally referred to as image or vector field (VF) segementation. Research using VF segementation extends back to the 1970s and was applied to large portions of the Illinois landscape in a cooperative study conducted by the ILNHS and the University of Illinois, Beckman Institute for Advanced Sciences (Kerfoot and Bresler 1993). It has been ascertained that VF segmentation is effective in discriminating homogeneous landscape elements within Landsat TM imagery. Extensive analysis using two TM full-scenes subjected to VF segmentation strongly indicates that unsupervised clustering and subsequent classification based upon image data is better compared to the same analyses using the original TM image data. In additon, it is anticipated that the application of VF segmentation will improve classification accuracy by minimizing the within-class variance.

Second Cut Classification

The CTAP vegetation map identified 19 broad land use classes in Illinois, covering urban areas, woodlands, grasslands, agricultural lands and wetlands. Using the natural cover delineations from the CTAP classification, a Boolean mask will be used to further classify the broad natural CTAP classes into community/alliance classes where applicable. A total of 140 spectral signatures for each Boolean masked area within each VF-segmented TM scene will be extracted utilizing an unsupervised isodata K-means clustering procedure (Duda and Hart 1973). Final unsupervised classification of each TM scene will be achieved from use of a maximum-likelihood classifier, which improves the classification accuracy over other classifiers (Gong and Howarth 1990). A pilot project has been completed on a portion of the Shawnee National Forest in southern Illinois using the methodology described above. An accuracy assessment will be conducted once spring leafout has occurred.

Additional Coverages

Boundaries for all federally- and state-owned lands has been completed. Attributing is nearly complete, and management status codes are currently being input and verified. The ILNHS has extensive vertebrate distribution records, and wildlife habitat relationship models are being developed for several test species. Distribution maps and occurrence records are currently being linked to the ILNHS home page.

Literature Cited

Duda, R.D., and P.E. Hart. 1973. Pattern recognition and scene analysis. J. Wiley and Sons, New Yourk. 482 pp.

Gong, P., and P.J. Howarth. 1990. An assessment of some factors influencing multispectral land cover classification. Photogrammetric Engineering and Remote Sensing 56:597- 603.

Kerfoot, I.B., and Y. Bresler. 1993. Design and analysis of an information theoretic algorithm for vector field segmentation. Proceedings, IS&T/SPIE Symposium of Electrical Engineering and Technology 1904:1-12.

Luman, D.D., and M. Ji. Accepted. The Lake Michigan ozone study: An application of satellite-based land use and land cover mapping to large-scale emissions inventory analysis. Photogrammetric Engineering and Remote Sensing.

Nichol, D.G. 1990. Region adjacency analysis of remotely-sensed imagery. International Journal of Remote Sensing 11:1089-2101.

Woodcock, C., and V.J. Harward. 1992. Nested-hierarchical scene models and image segmentation. International Journal of Remote Sensing 13:3167-3187.

Tony McKinney
Illinois Natural History Survey, Champaign, (217) 333-7022


The Indiana Gap Analysis Project is in the middle of project year two as of the end of 1995. Work during the second year has focused on the challenging task of developing a meaningful map of actual vegetation from TM imagery and available ancillary data. Considerable time was expended to establish a remote sensing methodology that will produce a defensible land cover classification. A useful preliminary classification for much of the forested southern part of the state has been produced. We have now begun to produce a final vegetation map of Indiana, using concurrent aerial photography interpretation, in conjunction with the Natural Resources Conservation Service and detailed ancillary data analysis.

Thanks to the support of the Indiana Department of Natural Resources (IDNR), Division of Fish and Wildlife, we have made significant progress on the development of vertebrate models for the 539 vertebrate species in the state. A preliminary methodology has been established to incorporate these data into ARC/INFO for analysis. We could complete this work by late summer of 1996. Agreements have been in place since year one for IDNR Natural Heritage data and managed areas data; revision of these data for use in Gap Analysis began this winter. Metadata protocols have been established and standardized across labs at Indiana University and Indiana State University.

Metaprojects (see feature article) have, as expected, manifested a variety of administrative and technical problems. The drive for client-oriented metaproject products, however, has indicated a weakness in the Indiana Gap Analysis methodology that we have attempted to strengthen. Our efforts to improve the methodology have focused on improving coordination among the principal GAP partners. Early metaprojects have begun to yield results. For example, the copperbelly water snake metaproject delivered hardcopy to the FWS in January. The landscape-scale wetland restoration project has produced preliminary products and reports and continues to generate interest among Indiana's conservation community. The Nature Conservancy metaproject at Blue River approaches completion. The TNC project at Pigeon River will become part of a larger TNC/IDNR/FWS cooperative study funded in part by EPA. Other metaprojects, including Population Viability Analysis, are ongoing. At least two new metaprojects appear to have funding and should go forward this spring. Discussions with the Indianapolis Zoo and the Indianapolis Children's Museum to establish a biodiversity education metaproject appear promising as does a proposal to evaluate the importance of agricultural landscapes to biodiversity.

We will carry out an expert review of land cover maps and vertebrate models this spring, which should guide us toward a final product. We will also design a formal accuracy assessment of the land cover map toward the end of project year two. Finally, the Indiana Gap Analysis Project will continue to pursue metaprojects as funding becomes available and as metaprojects are feasible with respect to producing basic Gap Analysis products.

Forest Clark
U.S. Fish and Wildlife Service, Bloomingto, (812) 334-4261/206


The Kansas Gap Analysis Project (KS-GAP) is in the early stages of map development. The primary cooperators involved are the Kansas Biological Survey (KBS) at the University of Kansas (KU) and the Geography Department at Kansas State University (KSU). Jack Cully of the Cooperative Fish and Wildlife Research Unit at KSU is coordinating these efforts. The Kansas Applied Remote Sensing Program of KBS began work on developing a prototype land cover layer in late 1995. A multitemporal classification approach involving three TM scenes (late spring, early summer, and late summer) will be used to identify natural land cover types in southwest Kansas. The goal is to map land cover to the alliance level using the modified TNC-UNESCO vegetation classification developed by KBS in cooperation with The Nature Conservancy's (TNC) Midwest Regional Office.

The Geography Department at KSU has begun work on tiling USGS 1:24,000 quadrangle maps across the state. The purpose is to create mylar overlays upon which protected land areas can be traced and then scanned into a GIS land management layer. Maps showing protected lands in Kansas are available from KBS, and the resulting GIS layer will serve to secure this data set in digital format. The quadrangle maps were originally used to develop a statewide soils map for the USDA Natural Resources Conservation Service. The soils map will provide a useful layer for the vertebrate distribution models and will facilitate identification of natural vegetation types.

KS-GAP recently hired Dr. Glennis Kaufman, who received her Ph.D. in Biology at KSU, as a 1/2 time coordinator. Dr. Kaufman, a long-time resident of Kansas, has particular expertise regarding vertebrate distributions in the state. She is also well-connected with other biologists and biological collections in Kansas. This year, her responsibilities will be focused on developing partnerships and developing support for the project. Beginning next year, she will become involved in developing the animal distribution layer.

Chris Lauver, Kansas Biological Survey
University of Kansas, Lawrence, (913) 864-7691

Jack Cully, KS Cooperative Fish and Wildlife Research Unit Kansas State University, Manhattan, (913) 532-6534


The Louisiana Gap Analysis Project is currently in its third fiscal year. The entire state has been divided into a grid of 332 unclassified cluster panels of 900 x 900 pixels each. Strategies to insure connectivity among classified TM panels and also between classified TM panels and National Wetland Inventory (NWI) data panels are being developed. Recently, members of the LA-GAP team completed an initial ground-truthing survey for post-visual classification of the vegetation map. Cognitive, or on-screen, classification of the land cover map was completed in late August 1995. The GAP team is currently compiling the ground-truth data into a database. This database, along with the NWI database, and the use of color infrared (CIR) aerial photography will be used to refine the visual interpretation of the TM data. The CIR photography is currently being indexed, scanned, and stored on CD-ROM. Another auxiliary data set that is being compiled is a TM/SPOT merge. These two auxiliary data sets will provide a means of performing a Classification Accuracy Assessment statement. Definitions to the land cover classification terms are in progress.

Three GAP meetings were held during 1995, involving cooperators and individuals interested in the GAP project. Two of the meetings took place at the NBS Southern Science Center and one at the Corps of Engineers' New Orleans office. Attending were representatives from Louisiana Natural Heritage Program, The Nature Conservancy, Louisiana Department of Natural Resources, Louisiana Department of Wildlife and Fisheries, Louisiana Department of Environmental Quality, University of Southwestern Louisiana, Louisiana State University, University of Northeastern Louisiana, University of Northwestern Louisiana, Tulane University, Loyola University, Corps of Engineers-New Orleans District, U. S. Fish and Wildlife Service, Environmental Protection Agency, U. S. Forest Service, Natural Resources Conservation Service, National Marine Fisheries Service, and U. S. Geological Survey.

Jimmy Johnston and Steve Hartley
Southern Science Center, Lafayette, (318) 266-8556


Phase I of Maine Gap Analysis is almost complete (awaiting the final habitat map from the University of Massachusetts). Phase II is well under way with the development of the habitat map using aerial videography and ancillary databases as training information. Dr. Steve Sader, a remote sensing specialist from the University of Maine, has joined as a co-principal investigator. Dr. Zhangshi Yin has joined the team as a research associate experienced in processing satellite data.

In phase II, 1993 TM imagery will be used to identify habitat types. During the upcoming months, all TM images will be converted to the same format, coordinates, and grid size. A mosaic of images, including a 10-km boundary around the state, will be developed from the 1993 imagery. Clouds will be masked out, and a principal component analysis of the six TM bands will be used for data reduction. Ultimately, supervised, unsupervised, and guided clustering algorithms will be used to classify habitats within individual ecoregions. These ecoregions will be stitched together, and the resulting habitat map will be tested using aerial videography and ground-truthing.

Aerial videography with the video frames positioned geographically using a Global Positioning System will be used to identify satellite image signatures. Videography transects totaling 7,100 km statewide were obtained in summer and fall 1994. Maine was divided into 8 regions, and 6 to 8 examples of each habitat within each region were ground-truthed. Habitats on videography were printed out, and 120 sites along public roadways were visited to check the relationship between videography and ground observations. A catalog of videography has been developed to use as reference in classifying satellite data and in testing the resulting map.

As part of Maine Gap Analysis and for use in other research, we have contracted for the acquisition of aerial videography along 48 Breeding Bird Survey routes. During fall 1995, 26 of these routes were flown; we anticipate completing the flights in 1996. This videography will be used to assess the accuracy of the predicted distributions of birds in Maine based on Gap Analysis.

Species synopses have been developed for each of the 278 terrestrial vertebrates that breed in Maine. The amphibian and reptile synopses are finalized, mammals have been reviewed and await final editing, and bird synopses are being reviewed. Synopses have been used to assist personnel of the Maine Forest Biodiversity Project and commercial forest industry personnel. After being finalized, the species synopses will be used in Randy Boone's doctoral research. He plans to use the range maps to research the effect of generalizing distributions to coarser political units (e.g., counties). Ultimately, they will be reformatted to be more concise and published in two volumes.

Efforts to develop scores for how well species should be predicted by Gap Analysis have expanded. We will be developing predictability scores for the species of Maine and selected western states where Gap Analyses have been completed. Predictability as assigned, using ecological variables, will be compared to species lists from conservation areas to test agreement. Should correlations be high, others conducting Gap Analysis will be able to judge a priori which species should require more effort during modeling.

A digital database of land ownership and an accompanying paper map were purchased from a local contractor. We are coordinating with the Maine State Planning Office to ensure that Maine conservation lands are accurately mapped and made current to 1993. During the upcoming months, we will classify public lands as to the level of biodiversity conservation they provide. We will be finalizing the species synopses that we have developed for Maine, and Randy Boone will be completing a thesis. We may meet with other GAP personnel late this winter to further the research on predictability scores for vertebrates.

William B. Krohn and Randall B. Boone
ME Cooperative Fish and Wildlife Research Unit
University of Maine, Orono, (207) 581-2870


(see Maryland, Delaware, and New Jersy)

Maryland, Delaware, and New Jersey

In 1995, the Mid-Atlantic Gap Analysis Project (MidA-GAP) finalized cooperative agreements with three museums to obtain data for vertebrate species modeling. Mammal data were also acquired from the University of Delaware's mammal collection. A Memorandum of Agreement (MOA) was completed with the Delaware Natural Heritage Program, and a complete copy of their Biological and Conservation Database (BCD) was obtained. As a result of this latest MOA, the MidA-GAP now has BCD data for its entire project area. A cooperative agreement was also entered into with the Birds of Delaware Publication Committee, and a complete digital copy of the Breeding Bird Atlas (BBA) data has been acquired. Some butterfly data have been obtained, including Opler's county-based data (Stanford and Opler 1993) and data from the University of Delaware, and other data sets have been identified.

Several GIS coverages have been developed or acquired, including a coverage of Delaware Natural Heritage element occurrence locations, as well as a previously developed element occurrence coverage for New Jersey. The Biodiversity Research Consortium (hexagon) project is under way in Maryland and Delaware, with several draft range maps completed (see Master and Jennings 1993). Preliminary GIS database structures have been designed for all vertebrate distribution coverages.

A University of Delaware graduate student is conducting a pilot project involving random sampling of vertebrates, including live trapping, in a variety of habitats within a small watershed in Delaware. A hand-held GPS receiver is being used to record precise geographic positions of occurrences and attribute data about habitat features. Preliminary field work has yielded some bird and amphibian data, some of which have been converted to GIS coverages. Another University of Delaware student, working as an intern, will be conducting small mammal live trapping in another watershed. The data from these projects will be used in accuracy assessment. Volunteers from the University of Delaware's Wildlife Program spent 60 hours in the field using standard field data forms to collect data on vertebrate breeding and associated habitat.

In Maryland, an NBS state partnership project is developing protocols for censusing reptiles and amphibians in different physiographic provinces of the region. The data collected during the study will be used for accuracy assessment. Breeding Bird Survey route stops are being digitized for Maryland in order to make use of the BBS data.

Expert reviewers have been found for the bird, herptile, butterfly, and bat models and distribution maps. Literature syntheses of habitat requirements have been completed, in a standardized format, for approximately 25 percent of all species to be modeled, and most of the remaining work will be completed this winter. MidA-GAP investigators in Delaware are involved in the development of a state desk-top mapping, database, and decision management system which will eventually include GAP data sets.

The air video project began in late fall after working through unexpected hardware problems. West Virginia GAP is conducting flights for MidA-GAP, and a second flight is planned for spring of 1996 after leaf-out. Video will be instrumental in developing the vegetation maps using protocols as set forth by Slaymaker (in press) and others.

The majority of the MRLC TM scenes have been received. These were re-registered after the registration accompanying the files was found to be off by more than 250 meters. All hyperclustered data received have been registered, and work has begun on these for use with the video. Preliminary land cover maps are expected to be completed in mid-summer 1996. MidA-GAP is working to collaborate on other projects in the region such as the NPS-TNC effort to map vegetation for national parks. Opportunities to do more of these projects are expected as we get farther along with the vegetation mapping process.

Literature Cited

Master, L., and M.D. Jennings. 1993. Hexagons: A new way to display predicted distributions of vertebrate species. GAP Analysis Bulletin No. 3:6-7.

Opler, P.A., compiler. 1994. County atlas of eastern United States butterflies. National Biological Service, 1201 Oak Ridge Drive, Suite 200, Ft. Collins, CO 80525.

Slaymaker, D., K. Jones, C. Griffin, and J. Finn. In press. Uses of aerial videography in Gap Analysis for deciduous forests in New England. In J.M. Scott, T.H. Tear, and F. Davis, editors. Gap Analysis: A landscape approach to biodiversity planning. American Society for Photogrammetry and Remote Sensing, Bethesda, Maryland.

Stanford, R.E., and P.A. Opler. 1993. Atlas of western USA butterflies including adjacent parts of Canada and Mexico. Denver and Fort Collins, CO.

Ann Rasberry
Maryland Department of Natural Resources, Annapolis, (410) 974-3195


(see Massachusetts, Connecticut, and Rhode Island)

Massachusetts/Connecticut/Rhode Island

The University of Massachusetts and the Massachusetts Cooperative Fish and Wildlife Research Unit are cooperating with the Vermont and Maine Cooperative Fish and Wildlife Research Units in the New England Gap Analysis Project. A primary focus of the Gap Analysis activities in Massachusetts has been development of a systematic approach for mapping deciduous forests. The New England landscape is 50 to 95% forested, with a wide variety of forest types occurring in relatively small stands interspersed throughout the region. These regional vegetation characteristics pose new challenges for developing an efficient and reliable methodology for developing base vegetation maps in New England and for much of the eastern deciduous forested region of the U.S.

Our approach has been to use hyperclustered, multitemporal Landsat TM imagery in combination with aerial videography. The MRLC program provided us with 12-band hyperclustered TM images that combined spring and summer coverages. Ground reference of vegetation cover was obtained from a grid of large scale GPS-logged videography flown over the region. Video data were collected along a 20 km-grid pattern using two Super 8 video recorders mounted on a Cessna 172. One video camera was set at wide angle, the other at 12x zoom, providing a swath of 30 m wide large-scale imagery down the middle of a 0.4 km wide-angle coverage when flying at 600 m above ground level. The GPS time code was recorded onto the video images and the audio track.

After developing a visual key of forest types obtained from video prints and field visits to training sites, the flight line was displayed over the hyperclustered image. The corresponding video images were used to label the vegetation at nearly 18,000 sample points from approximately 2,300 locations. Thirty Natural Community Alliances were identified. Through an iterative process, inference rules were developed and the image was classified. Accuracy was determined by an error matrix using a stratified subsample of video points that had been set aside during the video interpretation phase. The overall accuracy for all classes was nearly 90%.

We believe that the hyperclustered TM image represents a considerable improvement in the discrimination of spectral classes, especially in forested regions. Further, GPS-logged aerial videography provides a time- and cost-efficient method for obtaining sufficient samples of ground-truthed data to label the spectral classes in the TM scene. A measure of this methodology's usefulness is its applicability to other Gap Analysis projects. We have conducted training workshops, set up interpretation systems, or flown aerial videography for other GAP projects in 9 states. Regional workshops in the Northeast are ongoing to standardize video interpretation criteria, vegetation classification, and species habitat models.

The Massachusetts Gap Analysis team also continues to be involved with international initiatives in biodiversity inventory, cooperating with projects in Romania, Madagascar, Portugal, Ukraine, and Mexico. Our efforts center on providing technical tools and training to small groups of foreign scientists and development of GIS-based products that contribute to their conservation planning needs. Our goals focus on the rapid development of in-country GIS capabilities, making critical data available for resource management decisions and strengthening institutions within these host countries. The Gap Analysis approach is rapidly beginning to be integrated into conservation management programs throughout the world.

Curt Griffin, Dana Slaymaker, and Jack Finn
University of Massachusetts, Amherst, (413) 545-2640


(see Michigan, Minnesota, and Wisconsin)

Michigan, Minnesota, and Wisconsin

(Upper Midwest Gap Analysis Project)

The Upper Midwest Gap Analysis Program (UM-GAP) has received most of the Landsat Thematic Mapper (TM) scenes necessary for classification of actual vegetation of Michigan, Minnesota, and Wisconsin. All scenes were reviewed at the Environmental Management Technical Center (EMTC) for quality control and were converted to an ERDAS Imagine format before being distributed to state partners. The Departments of Natural Resources of Minnesota and Wisconsin are in the process of classifying scenes for their states. Michigan is soliciting contractual support to classify the northern half of the Lower Peninsula. The EMTC will retain responsibility for classification of the Upper Peninsula of Michigan and has recently begun that effort. UM-GAP coordination efforts now also include Illinois, and the EMTC is working with Indiana and Iowa to encourage regionally compatible vegetation classifications. A series of meetings to promote that effort was held this winter.

The U.S. Forest Service has contributed to the UM-GAP vegetation mapping effort by assisting in the acquisition of additional TM imagery for the Lower Peninsula of Michigan. The Forest Service's Great Lakes Assessment will benefit from the use of UM-GAP-developed GIS coverages of current vegetation and predicted species distribution. In addition, the North Central Forest Experiment Station has signed a Memorandum of Understanding with the EMTC to share TM imagery and Forest Inventory and Analysis (FIA) plot data. The Forest Service will use the imagery to georeference their FIA plots, and UM-GAP will use FIA plot information for accuracy assessment.

In an effort to develop a uniform, current vegetation map for the Upper Great Lakes Region of the United States, UM-GAP has developed a common image processing protocol and a common classification scheme for Michigan, Minnesota, and Wisconsin. The classification scheme was developed in accordance with National GAP standards, following The Nature Conservancy/UNESCO design. Dr. Thomas Lillesand, Director of the Environmental Remote Sensing Center, University of Wisconsin-Madison, developed the protocol in cooperation with the GEO Services Division of the Wisconsin Department of Natural Resources. Technical approaches of the protocol include (1) use of multidate TM scenes, (2) use of GIS-assisted preclassification stratification into urban/nonurban and upland/lowland categories, (3) use of an extendible classification scheme which can be cross-walked to other classification systems, (4) preclassification stratification of scenes into spectrally consistent geographic subscenes based on ecoregion boundaries, (5) use of guided clustering techniques for classification of nonurban uplands, and (6) use of geographically stratified, systematic, nonaligned sampling for collection of training and accuracy assessment data. UM-GAP also will be testing the aerial videography system acquired by the National GAP office for acquisition of training site and accuracy assessment data. The protocol, in a compressed Postscript format, can be downloaded from the EMTC's anonymous FTP site ( In an effort to coordinate TM scene classification among UM-GAP's three state partners, the EMTC has also established an e-mail technical discussion list ( By using the list to discuss TM scene processing issues, state partners share experiences in solving problems with corner coordinates, file headers, and software—saving much time, frustration, and duplication of effort.

UM-GAP has also established a home page on the World Wide Web (see The image processing protocol can also be retrieved directly from that page. A false color-infrared composite of the Landsat TM satellite imagery covering the Chippewa Plains Ecoregion Subsection in Minnesota is also available through the UM-GAP home page. The coverage is available as single-band ERDAS Imagine files, clipped to 1:100,000-scale USGS quadrangles. The files can be used as image backdrops in GIS programs, including ARC/INFO. These files have been used to assist in the delineation of land type associations, the ecoregion unit below the subsection level.

A unified regional effort to develop species-habitat associations and predicted vertebrate distributions is being coordinated by the EMTC with the University of Wisconsin-Madison providing technical assistance and oversight. A committee is currently being formed to oversee this effort, with membership including representatives of the U.S. Forest Service, the National Biological Service, and the Michigan, Minnesota, and Wisconsin Departments of Natural Resources. In addition, UM-GAP is working with Illinois, Indiana, and Iowa in exploring the potential for a larger regional effort to map predicted species distributions.

Daniel Fitzpatrick
Environmental Management Technical Center
National Biological Survey, Onalaska, Wisconsin, (608) 783-7550/12


(see Michigan, Minnesota, and Wisconsin)


Missouri GAP has been integrated into the Missouri Resource Assessment Partnership (MoRAP). In doing so, some of the timelines have had to be readjusted. The new MoRAP Director, Dr. David Diamond, is now on board and is coordinating the GAP program within MoRAP.

All of the TM imagery for the state was received in November and December 1995. We have purchased PCI image processing software which resides on both workstation and PC platforms. An additional 8 gigabytes of memory for these scenes and their analysis have also been bought, and the imagery has been loaded onto disk. The final land cover classification scheme is under development for these images. Once finalized, the classification process will be initiated on these scenes. Two pilot areas are being examined to investigate different protocols for the detailed classification.

The first iteration of the public lands database and vertebrate distribution mapping should be completed by the end of May. These will then be sent out for review. We are also in the process of creating a mechanism for the continual update of the public lands database with our cooperators. The socioeconomic database development is nearing completion for the state and will aid in the assessment of priorities for biodiversity mapping.

Tim Haithcoat
MO Cooperative Fish and Wildlife Research Unit
University of Missouri, Columbia, (573) 882-2324


The Montana GAP project began in the fall of 1991. Due to the size of the state relative to the amount of available data, much work remains to be done. Contingent on the availability of sufficient funding, we anticipate a completion date of December 1997.

To date, we have developed a 2-step digital process for classifying existing vegetation. The first step is discerning the pattern of spectral polygons and delimiting their boundaries. A classification algorithm, developed by Dr. Zhenkui Ma, accomplishes this by mimicking a TM false-color composite. The resulting unsupervised classification is then aggregated to a user-defined minimum mapping unit (MMU) using an object and rule-based merging algorithm. The second step entails a supervised classification to label the polygons. We have used these methods to map existing vegetation in western Montana at 2 ha MMU according to cover type, size class, and canopy closure. Forest Service field crews provided most of the ground-truth data used to train the supervised classification. We completed a land cover map of western Montana in March 1996 and began work on eastern Montana.

To fully utilize our detailed vegetation data, we intend to develop correspondingly detailed habitat models and species distribution maps and, in the process, build a wildlife habitat relationships (WHR) database specifically for Montana. Limited comparisons of habitat at 2 ha and 100 ha suggest that much could be gained by investing additional time and money in construction of a WHR executed at 2 ha MMU. However, given limited resources, we may opt to map distributions directly at the standard 100-ha MMU. Species distribution will be mapped by mid-1997.

The BLM has just finished a digital statewide land ownership layer, which we will recode to reflect management status as of the end of 1996. State cooperators include the Department of Fish, Wildlife, and Parks, Department of Natural Resources, Department of State Lands, State Library, and the University of Montana. Cooperating federal land management agencies include the Forest Service, Fish and Wildlife Service, Bureau of Land Management, and Natural Resource Conservation Service.

Roland Redmond
MT Cooperative Wildlife Research Unit
University of Montana, Missoula, (406) 243-4906


The Nebraska Gap Analysis Project was initiated in October 1995. Three graduate research assistants and one undergraduate assistant have been assigned to work on GAP. A full-time GAP coordinator was hired recently. Progress has been made in the following areas:

  1. A statewide mosaic of MRLC Landsat TM data is 85% complete. The mosaic and ancillary data sets have been co-registered in preparation for land cover analysis. The mosaic will also be used to prepare a poster to be printed by the Conservation and Survey Division, University of Nebraska-Lincoln.
  2. A cooperative agreement has been established with the State Museum of Natural History to automate faunal collection records, and data entry has begun.
  3. Nebraska GAP staff are working with the Nebraska Game and Parks Commission to acquire and augment digital data on land ownership and land management.
  4. An article on the Nebraska GAP project was published in the magazine Resource Notes, a publication of the Conservation and Survey Division, University of Nebraska-Lincoln.
  5. A cooperative agreement has been developed with the Natural Resources Conservation Service (NRCS) to cooperate on development of ancillary data sets. A formal request has been made to gain access to primary sampling unit data acquired for the NRI to facilitate GAP.

James W. Merchant and Marlen D. Eve
University of Nebraska, Lincoln, (402) 472-7531


The initial mapping of Nevada vegetation for Gap Analysis has been completed at the pixel level. There are 65 mapped cover-type classes statewide that have been identified. Accuracy assessment is currently under way. Once completed, the weighting matrix for vectorization of the raster layer into the GAP-specified 100-hectare MMU polygons will be developed. All animal models are completed and await the aggregation of the land cover map to the 100-ha MMU.

Thomas C. Edwards, Jr.
UT Cooperative Fish and Wildlife Research Unit
Utah State University, Logan, (801) 797-2529

New Hampshire

(see Vermont and New Hampshire)

New Jersey

(see Maryland, Delaware, and New Jersey)

New Mexico

Land Cover Mapping and Assessment

A land cover map comprised of 42 categories of vegetated and nonvegetated cover was completed during early 1995. We mapped 33 categories at the GAP standard of 100 ha MMU. Two vegetation communities were mapped at 2 ha and seven at 16 ha to ensure that their general distribution was not lost during aggregation to a larger MMU. This mapping effort resulted in a statewide map with approximately 26,000 mapped land cover polygons. We subsequently drew a stratified random sample representative of the mapped categories. Standardized polygon evaluation instructions, a data form, and location maps (with polygons numbered but unlabeled) were distributed to 43 assessment coordinators representing more than 100 assessment cooperators statewide. Final analysis and development of accuracy statements regarding the land cover map were pending at report time.

Animal Distribution

We are predicting the distribution of 602 vertebrate species based on associations with mapped land cover, watersheds, soils, elevation, precipitation, hydrology, and slope. Our basic approach was to develop a "hypercoverage" consisting of the intersection of all polygons in the previously listed themes. Each vertebrate species is then assigned a presence or absence value for each hypercoverage polygon based on an algorithm of theme associations in a database system. Preliminary distributions were predicted, graphed, and submitted to expert cooperators representing bird, mammal, and herpetozoan expertise statewide. Wildlife models were altered based on expert comment, the hypercoverage was updated to represent corrections needed in individual themes, and these revised maps were distributed at the end of 1995.

Management Categorization

Specific management descriptions and tract boundaries were received from a wide array of public resource management agencies and private interests statewide. These data were integrated with the previously obtained "Public Land Survey System and Ownership" data files compiled jointly by Bureau of Land Management, New Mexico Land Office, and the Earth Data Analysis Center at the University of New Mexico. The resultant land tract boundaries and descriptors were then converted to management status categories. We developed a dichotomous key to provide a repeatable method for assigning status categories. This is a draft process that was described in a poster presented at the Arkansas coordination meeting in July 1995. See page 20 for a more detailed description of the dichotomous key and opportunities for critique and improvement.


Analysis will follow the format described in the recent standard final report outline. In addition, NMGAP will explore the variation in avian richness between processes including and excluding wintering distribution. Overall project completion and reporting is expected by spring of 1996. Activities are under way to coordinate with the Earth Data Analysis Center in Albuquerque to be the in-state repository and distributor for the final digital files and metadata.

Bruce Thompson
NM Cooperative Fish and Wildlife Research Unit
New Mexico State University, Las Cruces, (505) 646-6093

New York

New York now has all the GIS coverages required for a state-level Gap Analysis of species either in hand or promised for delivery during 1996 by reliable cooperators. These coverages include breeding birds, mammals, reptiles, amphibians, butterflies, threatened/endangered/sensitive species, and public lands (state and federal, including large DOD holdings). The mammals database has been made available through cooperation with the New York State Museum and the USFWS Region 5 office. We continue to have excellent cooperation in developing the GAP database from our principal state cooperator, the New York State Department of Environmental Conservation (NYSDEC). The reptile and amphibian data are the most recent available from New York State, based on an atlas being produced by NYSDEC, and will be updated through 1996. Birds and butterflies are complete and linked to the GIS. We still are negotiating for additional, more up-to-date state forest boundaries. We also have compiled a digital elevation model for New York State, a model of growing degree days, digital soils information, and a compendium of published information about edaphic factors relating to vegetation types, all to assist us with developing and refining our vegetation classification.

We have completed a provisional, first-cut vegetation classification derived from Landsat TM imagery and using a single-scene clustering algorithm based on 100 clusters. The delivery of additional multitemporal, processed TM imagery from the EROS Data Center under our MRLC agreement, along with associated 240-cluster spectral data, was completed late in 1995. However, several problems need to be solved before we can use the EROS data to the fullest extent. Obtaining and installing Spectrum/Khoros software has been difficult, and some of the EROS-processed scenes were not accurately georeferenced.

In November 1995, representatives from the Pennsylvania, New York, and New England GAP projects met at the University of Vermont to discuss a land cover classification that would be appropriate for the Northeast. We incorporated elements of the modified UNESCO and TNC vegetation classification schemes into an expanded northeastern classification. We currently are reviewing the provisional vegetation classification scheme which expands upon both the modified UNESCO and TNC Heritage Program schemes, taking into account the extensive land use/land cover patterns that result largely from the activities of humans on the northeastern landscape of the United States. Additional meetings to discuss vegetation classification, vertebrate range delineation, and edge-matching are planned for 1996, prior to the national GAP PI meeting in Florida.

Charles R. Smith
Cornell University, Ithaca, (607) 255-3219

North Carolina

The current focus of the North Carolina Gap Analysis Project is the development of the vegetation data layer. We are nesting our vegetation datalayer into the landcover classification being done by the Multi-Resolution Land Characterization Consortium (MRLC). We are using the
non-vegetated classes from the MRLC classification as a masking tool and processing only those areas identified as natural vegetation. We have spent the past year gathering ancillary data, testing methodologies, and applying those methods to the classification vegetation for the Southern Atlantic Coastal Plain Flatwoods of North Carolina. Based on past experiences we knew field data would be a limiting factor to mapping vegetation at the alliance level. Our solution to this was to gather aerial videography data for areas known to be dominated by natural vegetation, use plant community data available from the North Carolina Natural Heritage Program, and get field ecologists into the computer lab to help develop an extensive point database of vegetation types. These points are then used to determine the correspondence between the alliances and the combinations of clustered Landsat TM imagery and ancillary datasets (i.e. National Wetlands Inventory and Natural Resources Conservation Service's Detailed County Soil Maps). We are in the process of summarizing the results of the preliminary mapping efforts with respect to the National Vegetation Classification. We will also be reviewing their potential with respect to ongoing conservation planning in the region.

In addition to the vegetation mapping, we have been developing cooperative relationships with agencies within the state, as well as with neighboring Gap Projects. Two of our cooperators, the North Carolina Heritage Program and the North Carolina State Museum of Natural Sciences are currently involved in a study titled "A Model Biodiversity Analysis for Southeastern North Carolina". Essentially, this is a mini-Gap Analysis. The vegetation datalayer we are developing will be an important contribution to this effort. We are in the process of developing an Memorandum of Understanding with the North Carolina Center for Geographic Information and Analysis, which serves as the state clearing house for geospatial data. A joint MOU between NC-GAP, The Natural Heritage Program and the North Carolina Wildlife Commission is also underway.

In this year we will continue the interpretation of videography and image processing for the northern coastal plain, as well as the piedmont of North Carolina. The mountains will be the focus for the 1998 field season. Vertebrate species range mapping and habitat modeling will begin in the southern coastal plain.

Alexa McKerrow - Project Coordinator
5123 Jordan Hall
Center for Earth Observation
Box 7106,
Raleigh, NC 27695-7106, (919) 515-3433
NC-GAP Homepage


During the second operational year of the Oklahoma Gap Analysis Project (OK-GAP), team members Mark Gregory and David Gade of Oklahoma State University and Mark Lomolino, Ian Butler, Dan Hough, and David Perault of the Oklahoma Biological Survey and University of Oklahoma have been working diligently on production of the vegetation, animal distribution, and land ownership and management data layers. This phase of the project has been challenging, mainly because of the size and complexity of the data and analyses. Needless to say, we are all anxious to print our first set of maps.

Land Cover Layer

A land cover classification scheme, prepared last year, is currently being modified to incorporate recent changes made by state botanists. So far, we have received 22 data sets of Thematic Mapper (TM) satellite data from USGS-EROS. These data sets cover 16 different scenes or locations: 11 scenes with data of a single date and five scenes with multitemporal data. In addition to the TM data, hyperclustered data of 6 and 12 channels have also been received for most of the scenes. Processing and preliminary analysis of the TM data have been initiated. We also received airborne videography data consisting of geocoded images from 17 north-south transects covering Oklahoma, flown in June and July 1994. These data will be used to interpret vegetative cover types and verify the TM analysis.

Animal Distribution Layer

The central database of vertebrates has been created and populated with species element codes, scientific names, common names, state and federal ranks and status, descriptions of habitat and environmental associations, and related information. Except for a few species, geographic ranges of mammals, birds, reptiles, and amphibians have been mapped, verified, and digitized. Locational databases are being compiled. Habitat associations have been encoded for all reptiles and amphibians and are being completed for birds and mammals. We are currently conducting a pilot study using a preliminary vegetation/land cover map to test procedures for overlaying vegetative cover, vertebrate distributions, and land ownership/management layers.

Land Ownership and Management Layer

We have digitized 379 public and private managed land units, including all 44 school land parcels. This represents about 95% of the public and private managed areas, open spaces, and wild lands in Oklahoma. The remaining major task will be to code each managed land unit using either the existing or revised national Gap Analysis land classification system. We will seek reviews of these code designations from individual landowners. As always, cooperators have played an important role in OK-GAP. For example, over 30 experts in the vegetation of Oklahoma reviewed and commented on our vegetation classification scheme. We plan to involve even more cooperators in the coming year to assist us in reviewing the first set of draft maps.

Bill Fisher
Oklahoma Cooperative Fish and Wildlife Research Unit
Oklahoma State University, Stillwater, (405) 744-6342


The Oregon actual vegetation map is being upgraded to current GAP standards. Twenty-three full or partial TM scenes cover the state. All scenes west of the crest of the Cascade Mountains are scheduled to be completed by April 1996. We anticipate the TM-based vegetation map will be completely classified and labeled by October 1996. An accuracy assessment is planned for 1997. Our land ownership and managed area data layers have also undergone minor updates. We are scheduled to complete 420 vertebrate distribution maps, based on vegetation cover type polygons and ancillary data (DEMs, hydrography) by June 1996.

We continue close cooperation with the Biodiversity Research Consortium (BRC). Several analyses of current Oregon data layers have already been conducted, including a preliminary gap analysis of 66 aggregated (to the alliance level) vegetation types. Five cover types were not represented in natural areas, and another 12 had less than 2% of their area in natural areas. These were mostly desert shrub and oak woodland cover types. Four of six types with more than 50% of their area in natural areas were high elevation forests and alpine communities. We are also cooperating with the U.S. Fish and Wildlife Service's Klamath Basin Ecosystem Restoration Strategy. In response to their needs, we have labeled our first TM scenes in the Klamath region and may embark on higher resolution mapping for the area.

In cooperation with researchers in Australia and the United Kingdom, we have used the BRC species distribution database to compare the efficiency and spatial outcomes of 19 reserve selection algorithms. Linear integer programming, only recently applied to this problem, provides optimal solutions to "cover the set" within a reasonable run time. Surprisingly, far simpler heuristic algorithms also perform very well. Most species (90%) are represented in five areas (EMAP 635 sq. km hexagons), but 23 areas are needed for complete coverage. We found that species peripheral to the state tended to guide the selection of the last dozen areas. This finding underscores the need to carry out bioregional analyses. A paper describing this research is in press in Biological Conservation.

Oregon GAP, the Oregon Department of Fish and Wildlife, and several other cooperators are beginning work under a $550,000 grant from EPA entitled, "Multiscale Biodiversity Conservation: A prototype process for Oregon." Part of the funding will support a graduate student in the Geosciences Department, Oregon State University, who will carry out more detailed gap analyses of our data sets. The Geosciences Department has recently created a new GIS lab called Terra Cognita (the name comes from our cooperator Ross Kiester) to support this type of research.

Much time and effort over the past year has gone into the preparation of a book-length manuscript, "Atlas of Oregon Wildlife." This is a natural spin-off of GAP species distribution maps and will make one of our products available to a general audience. Most reference works on Oregon vertebrates are over 50 years old, so this new synthesis of ecological and distributional information on Oregon species is a major contribution. We are negotiating to have it published by Oregon State University Press.

Blair Csuti
Idaho Cooperative Fish and Wildlife Research Unit
c/o U.S. Fish and Wildlife Service, Portland, (503) 231-6179

Thomas O'Neil
Oregon Department of Fish and Wildlife, Corvallis, (541) 757-4186


Extended Gap Analysis

Gap Analysis in Pennsylvania (PA-GAP) is viewed as the outset of a cooperative and comparative multiscale landscape information infrastructure initiative. Assessment of conservation status and opportunities for vertebrate habitats is only one among many motives for undertaking the initiative. The overarching goal of the initiative is one of landscape understanding.

Rather than choosing among various technological approaches to spatial information, we have conceived a progressive scenario of spatial technologies and information sources whereby analytical alternatives become mutually reinforcing. As the scenario proceeds from finer spatial scale to coarser scale, thematic content and landscape insight grow deeper. Thematic errors and/or uncertainties occurring earlier in the scenario can be redressed at later stages in different modes. Gap Analysis itself does not require that our full informational vision be realized, so the Gap Analysis time frame is one stage in a developmental odyssey of indefinite duration. Funding has been secured for a sequel to Gap Analysis that is concerned with statistical approaches to multiscale analysis of critical areas in watersheds and landscapes. The land cover portion of PA-GAP is described in detail in a feature article elsewhere in this bulletin (see page 14).

Biotic Occurrence Information

Digitized range maps have been superseded by The Nature Conservancy compilation that shows the level of evidence for species occurrence in each of the EPA 635 sq. km hexagons in Pennsylvania as part of a pilot project sponsored by the Biodiversity Research Consortium. These data sets have been recompiled to show species richness by hexagon for selected groups of taxa. We have, however, developed a new approach for representing and analyzing these data—as a surface having at least ordinal metrics. We refer to the results from this approach as "echelons." For example, the representation of species richness as a surface is one echelon.

An initial comparison is that of combined echelons for all vertebrates against separate echelons for mammals, birds, and fishes. These comparisons show obvious regional differences between the major groups of taxa. These differences are consistent with general knowledge of the respective taxa. The contrasts are sufficient to negate prospects of finding any single group of species that can serve to guide conservation work in general. From this initial comparison, we suggest that those conducting an analysis of conservation gaps revisit this issue carefully. The occurrences of birds were compiled from Pennsylvania Bird Atlas information into the EMAP hexagon grid. For the avian taxa, then, the hexagons provide a coarser scale view that is consistent with views for other taxa.

The Pennsylvania Game Commission has contributed its entire Fish and Wildlife Database to the PA-GAP effort. Occurrence information for this database is county-oriented. As with the EPA hexagons, the Pennsylvania Bird Atlas was a major source of information for this database as well. Awareness of overlap between databases is important for Gap Analysis and conservation work in general.

Land Management Information

Land ownership GIS data layers compiled by consultants for a low-level nuclear waste siting in the Commonwealth provide us with a point of departure. Some cooperators in Pennsylvania have expressed a desire for finer land management categories. The status of relatively small tracts is often of interest, particularly in a larger landscape context. To insure thorough consideration of alternatives and their merits, we have formed an internal task force on land management status. Meeting GAP standards is not an issue, but we want to accommodate the needs of cooperators and promote research. The best course may lead toward multiscale architectures accommodating incomplete information.

Habitat Models

This is one sphere in which our attempts to expand the knowledge envelope are very selective. We will seek consistency with New York and New England states relative to habitat models. Current New England work is being shared with Pennsylvania, and we have enlisted a new fisheries faculty member to fill a gap in our GAP team.

The prospective models are operable and respond primarily to generalized landscape variables. A lot is taken for granted with respect to habitat elements at finer scale. Apparent habitat as seen by these generalized models will provide a reference against which to compare more incisive models arising from research. More sophisticated habitat modeling is under way in related projects for bobcat and woodcock. As components of the multiscale landscape information infrastructure become available, they are being steadily incorporated into the advanced habitat modeling research.

Biodiversity/Conservation Planning

An initial biodiversity plan for Pennsylvania has been formulated by the Pennsylvania Biodiversity Technical Committee. Formation of a council to consider the recommendations of the plan is under way. These initiatives provide an immediate context for utilization of GAP results in Pennsylvania conservation programs.

Wayne Myers, Robert Brooks, Gerald Storm, and Joseph Bishop
Penn State University, University Park, (814) 865-8911

Rhode Island

(see Massachusetts, Connecticut, and Rhode Island)


Vegetation Mapping

The initial land cover map for the entire state was completed March 1995 using a hybrid unsupervised/supervised classification method in ERDAS Imagine. Twelve Landsat TM scenes from 1990 to 1993 were used to produce the Anderson Level II classification with an overall accuracy of 85%. Except for adding wetland data for just a few quadrangles, the land cover map of the state is completed. Refinement of the vegetation categories into a plant community-based map is under way using aerial videography.

The use of aerial videography for detailed mapping of the forest classes was adopted from Slaymaker et al. (in press). Approximately 4,600 km of video transects were flown over the forested lands in Tennessee during mid-April. Nearly 24 hours of video were gathered with two cameras providing continuous 0.5-km wide angle and 30-m zoom coverage. Almost 400 sites were visited throughout the summer and early fall. Interpretation of the video footage is currently under way. Video interpretation and additional classification of the TM scenes is being performed in states by physiographic province in order to take advantage of the variation across the state.

Interpreted video for the Mississippi Alluvial Plain and the Loess Plain of West Tennessee have been used to complete the labeling process of the unsupervised forest classification. Refinement of the spectral classification is being done to code confused spectral classes using information about the surrounding pixels (maximum, minimum, diversity, and majority values) as well as ancillary data (NWI, DEM, geology, soils, and buffered streams). The plant community classification developed by The Nature Conservancy's regional office is being used for general guidance. Limitations of our methodologies and differences in scale make it difficult to conform strictly to that classification.

Vertebrate Species Mapping

Distributions for the state's terrestrial vertebrate species were based on the county, physiographic province, and watershed of occurrence and then translated to the EPA-EMAP hexagonal grid. Range data from the Tennessee Animal Biographies System (TABS) and the Vertebrate Characterization Abstracts (VCA) were used to produce range maps for the 70 mammal, 55 reptile, and 65 amphibian species in the state. Range maps for 1,709 breeding birds were produced from the Tennessee Breeding Bird Atlas data, TABS, and VCA. Distributions of rare species are based on buffered point data provided by the Tennessee Natural Heritage Program. Reviews for rare species data will be done in cooperation with the authors of "Tennessee's Rare Wildlife." Reviews for the non-rare species will be conducted by state biologists.

Work on the habitat relationship models for west Tennessee has begun. The models will be cross-walked for each physiographic province as the vegetation classification becomes available. Data sources for the habitat model include TABS, VCA, and "The Land Manager's Guide to Birds of the South" (P. Hamel).

Land Ownership and Management

The public land coverage has been updated through a cooperative effort between the Tennessee Department of Environment and Conservation's Recreation Planning Division and the Tennessee Wildlife Resources Agency. A subcommittee of the Tennessee Biodiversity Team has categorized the majority of the lands as to their management status. Final revisions are scheduled for the spring.

Allen Gebhardt, Jeanette Jones, Susan Marden, Alexa McKerrow, and Patricia Miller
Tennessee Technological University/Tennessee Wildlife Resources Agency, Nashville, (615) 781-6555


Dr. Nancy Mathews, the Texas GAP Coordinator and Assistant Unit Leader for Wildlife left for a faculty position at the University of Wisconsin in August 1995. Raymond Sims was hired as the principal investigator of the Texas Gap Analysis Project (70% time) and Southwestern Regional Coordinator (30% time) in November. Raymond, a native Texan, comes to the GAP program from Blackstone, Virginia, where he was a GIS coordinator at Fort Pickett Military Reservation.

Land Cover Mapping

Land cover classification and mapping efforts are progressing. Texas A&M University has begun vegetation mapping for the eastern portion of the state, while Texas Tech University prepares to begin vegetation mapping for western Texas. A minimum mapping unit of 40 ha has been chosen for the entire vegetation map with an accuracy target of 80%. Vegetation classification will follow the UNESCO format with the addition of lower levels. To date we have received 29 TM images in raw format and 21 clustered TM images in Spectrum format. The total number of images we are slated to receive is 52. The projected completion date for the vegetation map is June 1998.

Because over 90% of Texas consists of privately owned lands, access for ground-truthing and accuracy assessment is limited. Airborne videography has provided a method of acquiring periodic georectified high-resolution images that can be utilized as a ground-truthing and accuracy assessment tool. SkyKing software, developed at Texas A&M's Mapping Sciences Lab (MSL), is utilized to interactively assign vegetation cover types to points within contiguous vegetated areas on georectified video frames. SkyKing writes files containing UTM points and the corresponding cover type. This file is then read into an enhanced version of Spectrum. Spectrum applies the cover types to all similar clusters in the hyperclustered TM files and reports errors. The final classified images are then aggregated up to the minimum mapping unit of 40 ha.

Texas A&M - Land Cover Map of East Texas

TX-GAP contracted with TX A&M's Mapping Sciences Lab (MSL) to map vegetation in the eastern portion of the state. Due to delays in receiving imagery, processing contracts, and utilizing Spectrum software, image processing is only now beginning. However, the MSL has accomplished several key goals including: collecting ancillary vegetation data, loading and compiling Spectrum 1.5, converting clustered data to Spectrum format, developing framework for state-mandated metadata files, checking and archiving TM raw and clustered imagery received to date, developing a vegetation photo key from airborne videography, and giving several presentations on current efforts.

Land Ownership

Of the federally managed lands in Texas, only national parks and U. S. Army Corps of Engineers lands have boundary data. Boundary data for state forests are complete, while state park and refuge boundary data are expected to be provided by Texas Parks and Wildlife Department.

Vertebrate Distributions

Experts in the various fields of Texas vertebrates have been contacted, and cooperators have been identified. Animal scientists and ecologists from universities and state and federal agencies are among cooperators identified thus far. Ancillary data is being collected, and mapping of vertebrate species distributions will commence in FY 96.

Dr. Nick Parker
Texas Cooperative Fish and Wildlife Research Unit
Texas Tech University, Lubbock, (806) 742-2851


Utah Gap Analysis was completed in spring 1995 and consists of a 1,138-page report plus two CD-ROMs containing all digital information. Effort has been focused on completing manuscripts and making presentations. Thus far, three manuscripts have been published, two are in press, and two in review. Eleven presentations have been made. Currently, work is emphasizing the optimal placement of reserves, given the existing reserved land, and assorted analyses of the pending wilderness bills for Utah.

Thomas C. Edwards, Jr.
UT Cooperative Fish and Wildlife Research Unit
Utah State University, Logan, (801) 797-2529


(see Vermont and New Hampshire)

Vermont and New Hampshire

After three years of compiling data for six states, the New England Gap Analysis Project has now subdivided the territory so that the Vermont Cooperative Fish and Wildlife Research Unit is coordinating the analysis for Vermont and New Hampshire. A major effort of the VT/NH project for the past year has been to gather and process aerial videography that is being used for a supervised classification of TM imagery for land cover maps of the two-state region. Transects have been flown over Vermont and New Hampshire during late fall, early spring, summer, and early fall 1994-1995. GPS codes, linked to time codes on the videotapes, have been corrected with base station data, then converted to ARC/INFO and ERDAS files. The quality of the videography is quite satisfactory and lends itself to detailed vegetation classification. We are concentrating on interpretation of videography for Vermont and classification of two scenes of TM imagery from the MRLC acquisition. We await additional TM scenes for New Hampshire.

Last October and November, separate meetings were held with cooperators in the two states. Both states have now initiated biodiversity projects that complement Gap Analysis. In Vermont, the Fish and Wildlife Department is funding a pilot project in the four southern counties that will bring Gap Analysis to a scale more useful for identifying conservation lands within the state, exclusive of a bioregional context. Private organizations, such as the Vermont Land Trust, also are cooperating. Initial plans were made at the cooperators' meetings for an assessment of the accuracy of predicting vertebrate distributions. Several cooperators have agreed to compile thorough lists of species in the hexagonal analysis units used in predicting species occurrence.

There have been some changes in personnel for the Vermont/New Hampshire GAP Project. In September, Ken Williams left the Vermont Cooperative Fish and Wildlife Unit, and David Capen assumed responsibility as Principal Investigator. Joel Schlagel, Systems Manager and Research Specialist, also left for a new position. Ernie Buford continues as a graduate research assistant, and is joined by Chris Boget, another graduate assistant, and Eric Lambert, a GIS and remote sensing specialist. Tom Allen, an assistant professor in Geography is leading the image processing effort. Some of our Gap Analysis products are displayed on the World Wide Web site for the Spatial Analysis Laboratory at

David Capen
VT Cooperative Fish and Wildlife Research Unit
University of Vermont, Burlington, (802) 656-3007


The Virginia Gap Analysis Project started in September 1994 and now involves three graduate students (Dave Morton, Scott Klopfer, and John McCombs), a part-time systems analyst (Blair Jones), a part-time GIS technician (Steve Phillips), and the principal investigator (Jeff Waldon). In addition to the main project, we have begun two associated projects with the Dept. of the Army at Fort Pickett and Fort A.P. Hill to collect verification data for vegetation and vertebrate distributions. We have received considerable support from cooperators, especially the Virginia Department of Game and Inland Fisheries. Two state coordination meetings were held, and our mailing list now contains over 60 biologists and land managers in Virginia. We participated in the Southern Appalachian Gap Analysis Coordination Meeting and agreed to cooperate with the other states in the Southern Appalachian Man and the Biosphere (SAMAB) region on classification consistency and edge-matching.

Hardware and Software

We have developed a network of Pentium-based microcomputers with multiple 1 gigabyte (gb) hard disks. One file server with 4 gb of on-line space and a 1.3-gb optical cartridge drive for near-line storage are available. Peripherals include an 8-mm tape drive and an HP 650C Designjet plotter. We have purchased one license of the Map and Image Processing System software (MIPS) for map composition and vector layer integration. Two licenses for EASI/PACE image processing software by PCI, Inc. were acquired for classification, rectification, and other processes.

Base and Validation Data

We have received and preprocessed all Landsat imagery available through MRLC and are pursuing additional scenes through SAMAB. All SPOT panchromatic imagery for Virginia was also received and preprocessed. We maintain complete copies of the 1:100k DLG transportation and hydrography layers, all available National Wetland Inventory Maps, the best available public lands layer, and a variety of other coverages. The Virginia Department of Game and Inland Fisheries is developing vertebrate distribution maps for all vertebrates and some invertebrates in Virginia. They are currently developing a review process and metadata guidelines for final map development.

Approximately 15 cover type maps from public lands in Virginia were received and processed. Nine of these areas were chosen as intensive test areas for vegetation model development. We just completed our test run of airborne videography and now have 19 hours of film to evaluate. We completed the first simple classified maps of study areas and are working through the process of model development.

Jeff Waldon
Fish and Wildlife Information Exchange, Blacksburg, (540) 231-7348


WA-GAP is nearing completion. The land cover map has been completed but has not been assessed for accuracy. All vertebrate data layers have been completed, and predicted distribution maps have been created for each species. The land ownership map has been completed. Analyses of the gaps in conservation for land cover types and vertebrate species have also been completed. A final report is in preparation and will be completed early in 1996.

Christian Grue
WA Cooperative Fish and Wildlife Research Unit
University of Washington, Seattle, (206) 543-6475

West Virginia

Our land cover mapping efforts are near completion in an intensive study area that was used for methods development (Allegheny Mountain Transitions subsection), and we are now proceeding with the rest of the state. Methods for videography in the mid-Appalachian area were developed, and West Virginia and some areas in contiguous states were filmed in October 1995. We plan to fly West Virginia again in May 1996 as part of a unified airborne video project covering MD, DE, NJ, VA, WV, and NC. The ecoregion map that we will use for stratifying the state was updated. We evaluated the utility of supervised vs. unsupervised classification for the Allegheny Mountain Transition subsection and found that a strategy of post-classification sorting, with elevation and slope/area index included as derived bands, is the most efficient use of the limited ground data we have for community alliances in this area. When the fall videography has been processed, we will proceed with a "hybrid" classification for the rest of the state. Ground vegetation surveys were completed during summer 1995; the data are being used to improve TNC alliances for West Virginia. Additional surveys will be conducted in 1996 for ground checks of the videography images. We have presented talks and training sessions on our methods for remote sensing to the Smithsonian Institution Conservation Research Center, at a U.S. Forest Service course on methods for ecosystem management, at the ACSM/ASPRS meeting in Charlotte, NC, in several classes at West Virginia University, and at an Earth Day booth.

The vertebrate species/habitat relationship database structure for West Virginia is complete. The database from the Virginia Department of Game and Inland Fisheries has been converted and merged with our database, and the data from DeGraaf et al. 1992 (New England wildlife: management of forested habitats) have been entered. Our wildlife database is based on the Society of American Foresters (SAF) forest cover types, The Nature Conservancy's (TNC) classification scheme (in progress), and the Cowardin et al. (1979) wetland classification scheme. We are using the SAF types because TNC's classification is not complete. As TNC data are made available, we will incorporate it into a cross-walk, so our database can be converted when all the TNC types are completed for West Virginia. The West Virginia Division of Natural Resources (WVDNR) and TNC are assisting with collecting locational data for the hexagons-of-occurrence for each species in the state. WVDNR has finished collecting the locational data on all species except butterflies and skippers.

Aquatic Diversity

Our goal is to inventory the level of aquatic diversity within certain drainages in West Virginia and to identify stream reaches or watersheds that offer high conservation potential. The initial study area includes the Monongahela and Potomac river basins, which together cover the northeastern third of the state. Data have been compiled for 94 watersheds within the study area for a number of environmental and human influence variables such as land use, elevation, bedrock geology, and population density. In addition, a fish collection database has been created to incorporate collection records from the West Virginia Division of Natural Resources, U.S. Forest Service, U.S. Environmental Protection Agency, U.S. Army Corps of Engineers, West Virginia University, and museum records into both the EPA River Reach coverage and the watershed coverage. Initial analysis of the watershed level data indicates that mining activities and limestone bedrock geology are related to fish species diversity at the watershed scale.

Sue Perry
WV Cooperative Fish and Wildlife Research Unit
West Virginia University, Morgantown, (304) 293-3794/2432


(see Michigan, Minnesota, Wisconsin)


The Wyoming Gap Analysis Project will submit a draft of its final report this summer. At this point, the three major WY-GAP databases (land cover, land ownership/status, and species distributions) are complete. The land cover database is already available on National GAP's WWW home page, and the other two databases will also be available on the Internet as soon as their documentation is complete.

Our recent efforts have been directed toward completing an expert review of the habitat associations and species range maps. We modeled the predicted distribution of 445 species in Wyoming using land cover, elevation, and riparian associations to create our final distribution maps. The review of these maps involved nearly 60 biologists/bird experts across the state. Several experts on a particular taxonomic group met to conduct the review to arrive at a general consensus on species distributions. In most cases, reviewers were satisfied with the maps but acknowledged that not enough is known about some species' distributions to produce a distribution map with much confidence. We are currently comparing our predicted species distributions with published species lists for several areas around the state to get a better idea of the amount of omission/commission error in our model predictions.

We have also just completed a state review of the assignment of protection status codes for which we used the key for categorization of land management developed by the New Mexico GAP project. We asked land managers from different state, federal, and private agencies to evaluate the protection status categories given to the lands under their jurisdictions. In most cases, the reviewers found the New Mexico key helpful in the categorization process. One exception occurred where mixed ownership and management objectives existed for the same area. For instance, private lands occurred within the boundaries of National Parks, Recreation Areas, and other management units. The private lands are not managed in the same manner as the federal lands and, as a result, were given a different protection status.

We expect to complete our Gap Analyses this spring and have a draft copy of our final report available by summer 1996 for review by our cooperators. Besides the three digital databases in ARC/INFO format and our final report, we will produce a book of range maps in the form of a black and white atlas, along with habitat associations and references for each species. In addition, a full-color land cover atlas for Wyoming has been developed, and opportunities for publication and distribution are being considered.

Upon completion of the WY-GAP project, the Wyoming Water Resources Center (WWRC) has been designated as the state repository for the WY-GAP databases. The WWRC's GIS lab is petitioning for the establishment of a data node under the National Biological Information Infrastructure (NBII). In December, the WWRC hosted a meeting of the WY-GAP cooperators to discuss objectives and priorities of the proposed data node. Sara Vickerman of the Oregon Biodiversity Project also discussed aspects of implementation of Gap Analysis to biodiversity at the state level and shared insights and recommendations for the implementation process in Wyoming.

Evelyn Merrill
WY Coop. Fish and Wildlife Research Unit
University of Wyoming, Laramie, (715) 346-4112

GAP Homepage - Table of Contents