GAP Analysis Program 2002 Meeting Program

Angermeier, Paul L. U.S. Geological Survey, Virginia Cooperative Fish and Wildlife Research Unit, Virginia Tech, Blacksburg, VA 24061-0321

Application of gap analysis to aquatic ecosystems marks a substantial shift in management perspective. A fundamental feature is the development of a georeferenced framework that facilitates linking a broad array of ecological and anthropogenic components over multiple spatial scales. This framework provides important new opportunities for identifying conservation targets, assessing threats to aquatic biodiversity, and catalyzing conservation actions. The data layers built for gap analyses will promote development of models that significantly improve our knowledge of distributions of rare species and communities. However, the dearth of certain key data types (e.g., flow regime) will also become more obvious. The spatial linkages inherent in gap analysis may help overcome the piecemeal approaches typically used to manage threats to aquatic ecosystems. But integrative assessments of threats will require new multidisciplinary thinking and spatially explicit modeling. The gaps between protected areas and areas with the most aquatic biodiversity are likely to be large and numerous, and not readily bridged by land acquisitions. Thus, aquatic gap analysis will underscore the need for innovative approaches to landscape-level conservation. The most important application of gap analysis products is their educational use in motivating society to care about native biota and functioning ecosystems, thereby bridging the huge gap in attitudes between conservationists and the general public.

Back to Top

THE APPROACH TO ANIMAL HABITAT MODELING IN THE SOUTHWEST REGIONAL GAP ANALYSIS PROJECT

Boykin, Kenneth G.1, Bruce C. Thompson2, Robert A. Deitner3, Frank La Sorte3, Scott Schrader5
1New Mexico Cooperative Fish and Wildlife Research Unit, New Mexico State University, P.O. Box 30003 MSC 4901, Las Cruces, NM 88003; (505) 646-6303; kboykin@nmsu.edu
2USGS-BRD, New Mexico Cooperative Fish and Wildlife Research Unit, New Mexico State University, P.O. Box 30003 MSC 4901, Las Cruces, NM 88003; (505) 646-6093; bthompso@nmsu.edu
3New Mexico Cooperative Fish and Wildlife Research Unit, New Mexico State University, P.O. Box 30003 MSC 4901, Las Cruces, NM 88003; (RAD) (505) 646-3355; bdeitner@nmsu.edu, (FS) (505) 646-3294; flasorte@nmsu.edu; (TSS) (505) 646-5022: schrader@nmsu.edu

Collecting and maintaining wildlife habitat relationship (WHR) model information is a significant task for every gap analysis project. Modification of these relationships through expert participation is another large task for state-based projects. The enormity of the task increases for regional projects that must deal with increased number of species, spatial variation of habitats, and spatial separation of project personnel and species experts. The Southwest Regional Gap Analysis Project is exploring ways that these model associations can be created, maintained, and refined to allow project personnel to input data, experts to help review and refine the models, and finally create a system that can be used by end users to modify the models. Our task of modeling 836 taxa within the 5-state region provides an excellent platform on which to test new modeling techniques and applications. We present our process in 1) identifying the taxa to be modeled in the region, 2) identifying lead states to pursue taxa WHR models, and 3) creation of a database in which to capture, review, and modify WHR associations.

GAP ANALYSIS IN VERMONT AND NEW HAMPSHIRE: PATTERNS OF VERTEBRATE DIVERSITY AND LANDSCAPE DIVERSITY

Capen, David E.1, and Ernest W. Buford2 1Spatial Analysis Laboratory, School of Natural Resources, University of Vermont, Burlington, VT 05405; (802) 656-3324; dcapen@snr.uvm.edu
2Spatial Analysis Laboratory, School of Natural Resources, University of Vermont, Burlington, VT 05405; (802) 656-3324; ebuford@snr.uvm.edu

The Vermont-New Hampshire Gap Analysis Project compiled data on distribution of 306 vertebrate species. Most of these data were related to town-based surveys, but many were point data, such as Breeding Bird Surveys. We used statistical ordination to identify gradients that were most obvious in delineating biological regions, then we drew a range map for each species based on boundaries of these regions. We mapped land cover/land use for the two states using 1992 TM imagery and extensive aerial videography. Nearly 11,000 polygons of stewardship land were digitized, but this accounts for only 18.6% of the area of Vermont and 21.2% of New Hampshire. As in many other states, only a small proportion of these lands are in categories 1 and 2 (3.1% and 8.1%, respectively). Most of the conserved lands are in high-elevation sites. Hotspots of vertebrate diversity, based on predicted habitat, are in the lowlands: the Champlain Valley of Vermont, the seacoast of New Hampshire, and the Connecticut River Valley. All are sites where human populations are most dense and where conservation is most needed. We also characterized landscape diversity in Vermont, using a 30 m grid comprised of four databases overlaid to create Landscape Diversity Units (LDUs). The four layers were bedrock geology, surficial deposits, landforms, and elevation zones. More than 1400 LDU codes were derived for Vermont. We used a "greedy" algorithm to account for the greatest richness of LDUs, and the greatest richness of predicted vertebrate distribution. The same hexagon was selected as the most rich for both measures. As we continued the efficient selection of landscape units, we quickly accounted for the distribution of vertebrates in the state. Eighty-eight percent of LDUs were captured with only 10 hexagonal sampling units, 20% of the area of the state. These units accounted for 97% of vertebrate distribution. This suggests that landscape diversity might be used as an efficient surrogate for vertebrate diversity.

ABUNDANCE AND NESTING SUCCESS OF ACADIAN FLYCATCHERS (Empidonax virescens) AT DIFFERENT SCALES OF HABITAT FRAGMENTATION

Chapa, Leonardo Illinois Natural History Survey, Center for Wildlife Ecology, 607 E. Peabody Dr., 172 NRB, Champaign, IL 61820; (217) 244-2211

To generate sound management strategies for some organisms such as neotropical migratory birds, presence/absence information within a given habitat may not be sufficient. Some species require large areas of "source" habitat that sustain populations in which reproduction may compensate for mortality. Because habitats are increasingly fragmented, small remaining patches may contain populations of neotropical migratory birds that are not successfully breeding due to increased nest predation and brood parasitism by Brown-headed Cowbirds (Molothrus ater). Extensive data exist on nesting success of birds for several states. Therefore, future approaches to animal modeling have the potential to include this information, as well as information related to habitat fragmentation. I examined the effects of habitat fragmentation at three spatial scales on nesting success (n = 1525 nests, 9 landscapes) and abundance (n = 4946 point counts) of a migratory songbird, the Acadian Flycatcher in Illinois. At the local scale, I compared nesting success at varying distances from and amounts of different edges (e.g., rivers, oxbow lakes, backwater swamps, roads, and agricultural). At the landscape scale, I compared abundance and nesting success among tracts of different sizes, shapes, and degrees of isolation. Finally, at the regional scale, I related nesting success to percent forest cover, average tract size, and percent forest interior. During my presentation I will present the results of this research, which was derived as an application of some of the gap analysis products in Illinois. Additionally, I will discuss some of the potential ways in which this information related to reproductive success and habitat fragmentation may be incorporated into the future generation of Gap Analysis projects.

GEOGRAPHIC AREA AND MINIMUM HABITAT THRESHOLDS IN RELATION TO ACCURACY ASSESSMENT OF PREDICTED VERTEBRATE DISTRIBUTIONS

Derting, Terry L.1, Adam Smith2, and Howard Whiteman3
1Department of Biological Sciences, 16th St., BL 334, Murray State University, Murray, KY 42071; (270) 762-6327; terry.derting@murraystate.edu
2Department of Biology, 1910 University Dr., Boise State University, Boise, ID 83725; (270) 426-3520; raptorbio@hotmail.com
3Department of Biological Sciences, 16th St., BL 334, Murray State University, Murray, KY 42071; (270) 762-6753; howard.whiteman@murraystate.edu

To assess the accuracy of the 361 predicted species' distributions produced through KY-GAP, we used three levels of assessment that differed from each other through the geographic size of the areas used for assessment. First, relatively small validation areas were assessed by measuring the omission error, commission error, and agreement of species checklists for natural areas around the state and species lists compiled along the North American Breeding Bird Survey (BBS) routes in Kentucky with our predicted species' distributions for those areas. Next, we conducted an assessment at the physiographic province level, using omission error and agreement rates within each of the eight provinces that occur in Kentucky. Lastly, we conducted an assessment based on independent locational records for the entire state, again calculating omission error and agreement rate for each species. The results of the assessments using checklists for small validation areas and point locations for the entire state indicated a high level of accuracy (( 80%) for most taxonomic groups. For the physiographic province assessment, agreement rates were very high and varied little among the provinces. The elevated degree of accuracy within this assessment was attributed to the tremendously large areas of the provinces. In order to further evaluate the importance of habitat quantity in our assessment of accuracy, we calculated the percentage of any given validation area predicted to provide habitat for each species. Three thresholds of predicted habitat (>1% habitat, >10% habitat, and >25% habitat) were tested for each validation area. Agreement rates (Nm) were compiled for species with > 1% habitat, >10% habitat, or > 25% of their habitat within each validation area. The agreement rates (Nm) were then compared with each other and with the original assessment (> 0% habitat; i.e., one 30 m2 pixel) at both the taxon and overall levels. Our results indicated that the validation areas were quite heterogeneous in terms of habitat types. Increases in the minimum habitat threshold resulted in exclusion of specific habitat types, causing significantly reduced agreement rates and increased omission errors.

A SIMULATED ANNEALING APPROACH TO SMOOTHING OUT THE GAPS

Dorfman, Dan1, and Pat Comer2 1Hawaii Natural Heritage Program-Gap Analysis Project, University of Hawaii, 3050 Maile Way, Gilmore #409, Honolulu, HI 96822 2NatureServe, 2060 Broadway, Suite 230, Boulder, CO 80302

The Hawaii conservation community has long been in search of a unified front in defense of Hawaiian biota. Virtually the entire ecosystem is at risk, and conservation organizations are rightly hesitant to throw out any of the parts. We present a simulated annealing algorithm approach to establishing a comprehensive conservation portfolio. This approach steps away from trying to fill gaps and instead looks for the most effective overall approach to targeting conservation resources. It is based on establishing an objective function that can represent multiple biodiversity goals while minimizing exposure to threat and respecting spatial distribution. This process allows us to establish conservation priorities through an objective repeatable process. The presentation will cover an approach based on the SITES tools developed for The Nature Conservancy by the National Center for Ecological Analysis and Synthesis. We will highlight use of terrestrial ecological systems, Ecological Land Units/biophysical habitat modeling, and the development of a suitability index for tracking ecological integrity. The SITES tool allows us to select between a heuristic or stepwise function and a simulated annealing comprehensive approach to priority setting. The SITES tool is available to GAP partners free of charge.

RESPONDING TO THE NEEDS OF NATURAL RESOURCE MANAGERS THROUGH THE NATUREMAPPING PROGRAM

Dvornich, Karen M. Washington Cooperative Fish and Wildlife Research Unit, University of Washington, Box 355020, Seattle, WA 98195-5020, (206) 616-2031; vicon@u.washington.edu, www.fish.washington.edu/naturemapping

The NatureMapping Program began in Washington State in 1993 and is spreading across the US. Patterns have emerged as it moves to other states. The demand by the public and the involvement of local, state, and federal agencies increased substantially by the end of the third year. The effort to educate and train the public continues as the dissemination and application of GAP and other biodiversity data sets become incorporated into training workshops and projects involving local, state, and federal agencies. Project CAT (Cougars and Teaching) is an eight-year study of the cougar population, prey base, and human development. The Mule Deer Project is studying the mule deer population and conducting vegetation analyses across the mid and eastern portion of Washington State and into Idaho. The Pierce County Biodiversity Management Areas Applications involves ground-truthing, habitat assessment and monitoring throughout biodiversity management areas derived from Washington Gap Analysis and other biodiversity data sets. In each of these projects, The NatureMapping Program is playing an important role to dissect the research questions into projects that K-12 schools and communities can conduct, thus involving the community in research and natural resource management issues. Credibility of data has been a primary concern. NatureMapping identified stumbling blocks the public faces while learning how to observe wildlife and how to collect, review, and analyze their data. Multilevel data collection training and analysis workshops have been developed. Computer data entry software, on-line data entry, and spreadsheet templates have been created. Personal Digital Assistants (PDA) loaded with a NatureMapping field notebook "sequence" will allow users to collect consistent data. GPS units and digital cameras attached to the PDA add to the credibility of data as well as the use of DRGs on the users' PCs to review before exporting data to ArcView or data submission to The NatureMapping Program.

PROGRESS AND FUTURE DIRECTION OF THE AQUATIC GAP PROJECT IN KANSAS

Gido, Keith1 , Jessica Kemp1, Bob Oakes1, Walter Dodds1, and Chris Guy2
1Divison of Biology, Kansas State University, Manhattan, KS; kgido@ksu.edu
2U.S. Geological Survey, Biological Resources Division, Kansas Cooperative Fish and Wildlife Research Unit

As part of the regional effort to develop an Aquatic GAP project for the lower Missouri River basin, we have been working with area partners in the state of Kansas. This presentation will focus on our recent progress. To date, we have held several meetings with local stakeholders and have identified specific uses for Aquatic GAP and available data sources for this region. Stakeholders have indicated they would like to use Aquatic GAP for a variety of management applications. These range from identifying streams with promising sport fishing opportunities to evaluating critical habitat for endangered species. We also have started to compile data on mussel and fish distributions in the state. Our primary biological data comes from stream surveys conducted by the Kansas Department of Wildlife and Parks, which includes over 800 collection sites sampled between 1995 and 2002. These data have already been used in association with GIS layers in the state to evaluate long-term changes (> 50 years) in fish assemblages in the Big Blue River system. Finally, we have begun two research projects to evaluate the effects of riparian buffers and reservoirs on biotic communities using data and GIS layers gathered for the Kansas Aquatic GAP project.

MODELING RANGE DISTRIBUTIONS OF TERRESTRIAL VERTEBRATES FROM SPECIES OCCURRENCES AND LANDSCAPE VARIABLES

Henebry, Geoffrey M.1, Brian C. Putz2, and James W. Merchant3
1CALMIT, University of Nebraska-Lincoln, 113 Nebraska Hall, Lincoln, NE 68588-0517; (402) 472-6158; ghenebry@calmit.unl.edu
2CALMIT, University of Nebraska-Lincoln, 113 Nebraska Hall, Lincoln, NE 68588-0517; (402) 472-2565; bputz@calmit.unl.edu
3CALMIT, University of Nebraska-Lincoln, 113 Nebraska Hall, Lincoln, NE 68588-0517; (402) 472-7531; jmerchant1@unl.edu

In order to provide a transparent and durable modeling framework for the range distributions of vertebrate species, the Nebraska Gap Analysis project has used recursive partitioning to develop "objective" semi-empirical models. Recursive partitioning algorithms predict membership of individual cases in classes of a categorical dependent variable from measurements of one or several independent variables. The motivation for using this strategy is two-fold: the resulting trees of decision points and values that form the models are readily understandable, debatable, and tunable; and the nonparametric modeling handles the multimodality common to regional species occurrence data. Although the best-known recursive partitioning algorithm is CART (Classification and Regression Trees), we have used QUEST (Quick, Unbiased, and Efficient Statistical Trees), a recent improvement on CART which greatly speeds up searching of the data space and which is more robust in the face of categorical variables with many levels. Species occurrence data were Breeding Bird Survey (BBS) route level composites since 1969 for the avifauna and georeferenced voucher specimens from the Nebraska State Museum collected since 1969. Explanatory factors included land cover class/vegetation alliance composition; surficial soils characteristics; climatic means, variance, and extremes; and terrain data. We will illustrate our modeling procedure, provide the model trees and resulting range distributions for representative species, and discuss the weaknesses and strengths of the framework.

AQUATIC GAP AND DECISION MAKING IN THE TALLAPOOSA RIVER BASIN, ALABAMA AND GEORGIA

Irwin, Elise1, James T. Peterson2, Byron J. Freeman3, Mary C. Freeman4, and Liz Kramer5
1USGS Alabama Cooperative Fish and Wildlife Research Unit, Auburn University, Auburn, Alabama, eirwin@acesag.auburn.edu
2USGS Georgia Cooperative Fish and Wildlife Research Unit, Warnell School of Forest Resources, University of Georgia, Athens, peterson@smokey.forestry.uga.edu
Institute of Ecology, University of Georgia, Athens, bud@ttrout.ecology.uga.edu
4USGS Patuxent Wildlife Research Unit, University of Georgia, mary_freeman@usgs.gov
5Institute of Ecology, University of Georgia, Athens, lkramer@arches.uga.edu

Methods for estimating species distributions and classifying stream segments for conservation and restoration have been developed by The Nature Conservancy and MORAP. These methods, however, do not explicitly quantify and incorporate uncertainty nor do they take into account incomplete detectability of species. Hence, their use is limited during formal quantitative decision making, a process that allows managers to estimate the potential consequences of each course of action (e.g., conservation strategies), select the most desirable alternative, and prioritize future research and monitoring efforts. We propose alternate approaches to modeling and predicting aquatic species distribution in relation to landscape features at various scales. We illustrate each approach using hypothetical management decisions and empirical species distribution and landscape data from the Tallapoosa River Basin, Alabama and Georgia.

ORGANIZING THE UPPER MISSOURI AQUATIC GAP PROJECT

Jenks, Jonathan A.1, Chad J. Kopplin1, Steven S. Wall1, Charles R. Berry, Jr.2, and Vickie J. Smith1
1Department of Wildlife and Fisheries Sciences, South Dakota State University, Brookings, SD
2U.S. Geological Survey (BRD), South Dakota Cooperative Fish and Wildlife Research Unit, South Dakota State University, Brookings, SD

The Upper Missouri River Basin (UMRB) includes 23 major drainages and six ecoregions, encompassing portions of six states and two Canadian provinces. For a gap analysis of this magnitude we followed four steps: 1) coordination across state and international borders, 2) attributing physical features to streams, 3) modeling fish distributions, and 4) conducting the gap analysis. An aquatic gap analysis was conducted as part of terrestrial GAP for South Dakota using the methods developed by the Missouri Resource Assessment Partnership (MoRAP) with some modifications. Coverages developed for South Dakota formed a base for data needs for other states in the UMRB. Over 60% of the 150 fish species found within the UMRB are common to South Dakota, and thus much data was already in hand at the start of the project. We coordinated with state, provincial, federal, and private agencies in North Dakota, Wyoming, Montana, Minnesota, Iowa, Alberta, and Saskatchewan to inform them of the project and to determine if data sets needed for our analyses were available. For areas where gap analysis was not completed (e.g., Canada) or GIS coverages were unavailable, we are producing the necessary GIS layers. To date we have attributed streams with seven physical habitat features (i.e., Strahler order, Shreve link, downstream link, stream size, size discrepancy, gradient, elevation) across the entire region. Fish distributions have been completed within the UMRB for Wyoming, Montana, North Dakota, and South Dakota. We will be intensively sampling an eight-digit hydrological unit from each state and province to fill voids in fish location data and verify our models predicting fish distributions. Sampling sites were chosen based on meetings with state and federal personnel familiar with the aquatic resources of those basins. Our success in organizing the UMRB Gap Analysis Project has been due to the input and cooperation of numerous people and agencies throughout the region.

BIODIVERSITY PREDICTIONS: INTEGRATING URBAN GROWTH MODELS WITH LAND COVER DATA AND SPECIES HABITAT INFORMATION

Jennings, Michael D. 1, presenting for Christopher B. Cogan2
1USGS Gap Analysis Program, Moscow, Idaho
2Alfred Wegener Institute, Bremerhaven, Germany

Habitat loss and subsequent fragmentation due to urban development are part of a larger suite of anthropogenic impacts on biodiversity, but they now rank among the principal causes of species endangerment in the United States. Several types of urban growth simulation models have been developed which can supply useful information for biodiversity planning. In many cases, however, the data required for biodiversity planning may not be compatible with the urban models, leading to analytical inaccuracies and misleading conclusions. Here, I briefly introduce a case study for biodiversity analysis and examine several lines of logic likely to be employed in such assessments. I conclude with a discussion of assumptions built into the data and their influence on model outcome.

ARE GAP DATA APPLICABLE TO PRODUCING EDUCATIONAL PRODUCTS?

Kaufman, Glennis A.1, Ryan L. Rehmeier2, and Dawn M. Kaufman3
1Division of Biology, Kansas State University, Manhattan, KS, gkaufman@ksu.edu
2Division of Biology, Kansas State University, Manhattan, KS, ryman@ksu.edu
3Division of Biology, Kansas State University, Manhattan, KS, dkaufman@ksu.edu

In Kansas, as in other states east of the Rocky Mountains, land ownership is mostly private (>98% of land in Kansas is privately owned). This has major consequences for the delivery of conservation practices such as GAP. Because of the lack of major land holdings by federal and state agencies, it is clear that a different strategy is required for conservation efforts to be effective. That is, education will be needed as a tool to convince private landowners that wildlife resources are part of our national heritage and that wildlife resources are worth conserving, both to individual people and to society as a whole. This education is best delivered through elementary students and interested adults. The KS-GAP Education Project will test the feasibility of developing education products for elementary schools from data collected by GAP state projects (KS-GAP data will serve as the basis of a "prototype" program). This project will assess the feasibility of forming partnerships with "trial" schools to deliver conservation products for use by elementary students. We will use these "trial" schools to help us develop and subsequently test classroom materials. The KS-GAP vertebrate database developed in KS-GAP Phase I will be changed from an input system to a user-friendly interactive output system that will serve as a discovery tool for students in the 6th grade.

APPLICATION OF INDEX OVERLAY AND FUZZY SET CARTOGRAPHIC MODELING TECHNIQUES TO GAP ANALYSIS

La Sorte, Frank A., Robert A. Deitner, T. Scott Schrader, Kenneth G. Boykin, and Bruce C. Thompson New Mexico Cooperative Fish and Wildlife Research Unit, P.O. Box 30003 MSC 4901, New Mexico State University, Las Cruces, NM 88003; (505) 646-3294

The gap analysis program has relied upon basic cartographic modeling techniques, specifically overlay models, to represent a species range and habitat relationships. These techniques are subjective, in that they are primarily knowledge driven, and non-probabilistic, in that input and output maps are binary (true or false) and lack a representation of ambiguity. For the present, these models will remain subjective because of the difficulty in acquiring detailed data in sufficient quantities to model multiple species objectively. For example, the Southwest Regional Gap Analysis Project (SWReGAP) is modeling 836 taxa, the majority of which lack detailed habitat information. However, with the introduction of slightly refined modeling techniques, binary representations can be modified to incorporate uncertainty. For example, methods based upon index overlay and fuzzy set theory incorporate subjective measures of uncertainty. These two techniques have seen extensive applications in GIS and are well suited to gap analysis. With inclusion of imprecision into fundamentally vague habitat models, we will provide a more realistic product that will receive greater acceptance and application. Retention of artificial binary products can sacrifice scientific merit and possibly promote misapplication of models.

BIOGEOGRAPHY, CONSERVATION BIOLOGY, AND THE WALLACEAN SHORTFALL

Lomolino, Mark V. Department of Environmental and Forest Biology, SUNY College of Environmental Science and Forestry, Syracuse, NY 13210; (315) 470-6805; island@esf.edu

"It is clear, therefore, that we are now in an altogether exceptional period of earth's history. We live in a zoologically impoverished world, from which all the hugest, and fiercest, and strangest forms have recently disappeared." This quotation is from Alfred Russel Wallace's Geographical Distribution of Animals (1876:150). Even then, long before evolutionary biology, ecology, and biogeography became established as respected disciplines, Wallace (1876) discussed what we would eventually refer to as the geography of extinction: “On a continent, the process of extinction will generally take effect on the circumference of the area of distribution, because it is there that the species comes into contact with such adverse conditions or competing forms as prevent it from advancing further." Today, conservation biologists are faced with an insidious catch-22: we need to know the most about the species that are most difficult to study-rarest, most isolated, and geographically most restricted species. The key to understanding and conserving biological diversity long into the future is a thorough knowledge of the geography of imperiled species. The critical paucity of such information-the Wallacean shortfall-remains one of the most important challenges for both biogeographers and conservation biologists.

MAPPING AN ECOLOGICAL SYSTEMS CLASSIFICATION: TESTING CONCEPTS

Manis, Gerald1, Keith Schulz2, John Lowry1, R. Douglas Ramsey1
1RS/GIS Laboratory, Utah State University, 5275 Old Main Hill, Logan, UT 84322
2NatureServe (formerly Association for Biodiversity Information) 2060 Broadway, Suite 230, Boulder, CO 80302

The Southwest Regional GAP project has focused preliminary land cover classification efforts on the alliance level of the National Vegetation Classification System (NVCS). While this goal has merits, accurately classifying at the alliance level is problematic and may be unachievable. In this presentation we broaden our concepts of the land cover mapping unit by grouping very similar alliances and spatially complex landscape associations of alliances into the "ecological system" naming convention proposed by The Nature Conservancy as a coarse filter for ecological community classification. Our objective is to identify a land cover mapping unit that is useful to land managers and is achievable given current methods, data, and technologies available for land cover mapping. We will examine the problems and merits of alliance versus ecological system labels for broad-scale mapping. We will look at how the ecological system classification was modeled and the degree of plant community and landscape separation achieved. We will compare the levels of mapping accuracy and community separation of what may be expected with ecological system level versus alliance level mapping.

OPPORTUNITIES FOR AND BARRIERS TO GAP IMPLEMENTATION

McClafferty, Julie A., and Jefferson L. Waldon Conservation Management Institute, Virginia Polytechnic Institute and State University, 203 W Roanoke Street, Blacksburg, VA 24061-0534

Gap Analysis products, including land cover, predicted species distributions, species richness indices, land stewardship maps, species habitat models, and even the GAP approach itself, are available, at least in draft form, for most states and regions of the country. In an informational sense, Gap Analysis is meeting its objectives admirably. Yet very few conservation agencies are taking full advantage of the available data and protocols. In this project we are exploring this issue through a series of mail surveys and focus group meetings involving current and potential GAP data users from various conservation agencies (public and private, profit and nonprofit). Our goals are to: 1) ascertain current awareness levels about Gap Analysis and its processes, products, and utility, 2) compare user needs and desires to the GAP concept and data capabilities, 3) identify technological capabilities/limitations of different types of decision makers, and 4) identify additional opportunities for GAP data integration. The results and recommendations from this study can be incorporated in GAP projects, so that products and processes are more compatible with user needs and ready for incorporation in decision-making processes with minimal "in-house" processing. This study also provides an educational forum for participants to learn about GAP data and its use. In this presentation we will present the results and implications of the mail survey and focus group meetings that have taken place to date.

REGIONAL GAP ANALYSIS AND RESEARCH IN GREAT LAKES TRIBUTARY RIVERS AND STREAMS AND COASTAL HABITATS

McKenna, James E., Jr.1, Steve Aichele2, Chris Castiglione3, John Gannon4, Kurt P. Kowalski4, Donna Myers5, Dora Passino-Reader4, Donald Schloesser4, and Jana Stewart6
1USGS/Great Lakes Science Center, Tunison Laboratory of Aquatic Science, Cortland, NY
2USGS/WRD, Michigan District, Lansing, MI
3U.S. Fish and Wildlife Service, Lower Great Lakes Fisheries Resources Office, Amherst, NY
4USGS/Great Lakes Science Center, Ann Arbor, MI
5USGS/Eastern Region, Columbus, OH 6USGS/WRD, Wisconsin District, Madison, WI

Biological diversity and the natural habitats supporting it are under attack, both globally and locally, from a variety of pressures, including habitat loss from urban expansion and resource exploitation and habitat degradation due to agricultural and industrial practices. The Great Lakes are the largest system of fresh water on earth and provide habitat for a variety of aquatic organisms, some unique to this system. They also include some of the most disturbed and depleted natural assemblages of aquatic organisms. The goals of GAP to conduct a large-scale spatial examination of habitat and species distributions and to identify gaps in those distributions, their protection, or knowledge, fit well with the needs of this region to better understand and manage its aquatic communities. The Great Lakes Aquatic GAP project is designed to classify rivers and streams and Great Lakes coastal areas, including islands, within the basin based on ecological conditions, gather aquatic biota distribution data, and link species distributions to the classified habitats. A geographic information system (GIS) and relational databases are used as tools for this analysis to map, summarize, characterize, and assess the status of aquatic biodiversity and associated conservation and knowledge gaps for the Great Lakes region. Progress toward completion of this project is staged according to the status and availability of aquatic habitat and biotic data in each state. Habitat classification and modeling will begin in Ohio, Michigan, and Wisconsin, followed by New York and then the other states bordering the Lakes. A pilot project to conduct gap analysis for coastal areas of the Great Lakes is being developed in parallel with the riverine component; the initial focus will be on an effective classification framework for coastal habitats throughout the Great Lakes. Methods used in the Great Lakes Aquatic GAP project will build upon earlier work in Missouri and other states. Standardization of methods and development of a regional database will ensure that results are comparable across the regional landscape. Data availability and applicability will be enhanced by this regional approach and hosting of the data sets on a centralized server.

THE DEVELOPMENT AND FUTURE APPLICATIONS OF A SPECIES-HABITAT AFFINITY DATABASE FOR THE STATEWIDE AQUATIC GAP ANALYSIS PROJECT IN VIRGINIA

Miller, Shelly, and Kevin Gooss Virginia Dept. of Game and Inland Fisheries, Wildlife Diversity Division, Richmond, VA; smiller@dgif.state.va.us

An integral part of the aquatic component of GAP is the development of a species-habitat affinity database. This database provides a link between stream reach classifications, described habitat preferences, and documented species' occurrences. Data from the VA Dept. of Game and Inland Fisheries' existing information system, known as the Virginia Fish and Wildlife Information Service (VAFWIS), has been migrated into an MS Access(tm) platform. Data are organized into four related tables: 1) a list of all fish, mollusk, and crayfish species known to occur in the state, 2) a list of georeferenced collection sites and associated species, including the associated reach identification number as assigned in the National Hydrography Dataset (NHD), 3) habitat preferences for each species derived from the existing VAFWIS database and a review of current literature, and 4) a list of all NHD stream reaches and associated habitat classifications. A crosswalk between habitat preference codes in the preexisting VAFWIS database and in the GAP species-habitat affinity database was created to facilitate populating the GAP database and to provide a means to update the VAFWIS database following completion of Aquatic GAP. The implementation of Aquatic GAP has come at a pivotal time for our agency, as it is embarking on a process of reengineering the VAFWIS to enhance functionality, streamline data entry, and provide greater GIS capabilities. The VAFWIS provides species and habitat data to a wide variety of user groups including agency biologists, researchers, consulting firms, other governmental agencies, nongovernmental organizations, and the general public. Integration of the data produced by Aquatic GAP into the reengineered system will enhance the ability of the VAFWIS to meet the needs of its users and to provide the data necessary to more effectively manage Virginia's aquatic resources.

STATUS OF LAND COVER MAPPING FOR THE SOUTHWEST GAP ANALYSIS PROJECT

Ramsey, R. Douglas, John Lowry, Gerald Manis, Todd Sajwaj, Lisa Langs, and Wendy Rieth Remote Sensing and GIS Laboratories, Department of Geography and Earth Resources, Utah State University, Logan Utah 84322-5240; (435) 797-3783; dougr@cnr.usu.edu

The Southwest Gap Analysis Project is the collaborative effort of the five southwestern states (Arizona, Colorado, Nevada, New Mexico, and Utah) to develop a regionwide gap analysis independent of state boundaries. Land cover mapping is being guided by Utah State University and the USGS EROS Data Center. This effort focuses on the development of standard image processing and classification techniques to be used throughout the five states. The goal is to generate a seamless land cover map depicting ecological systems and alliances as defined by the National Vegetation Classification System defined by The Nature Conservancy and NatureServe. Progress to date is the development of standard image preprocessing and classification methodology, a standard field data collection scheme, and ecologically significant and interpretive ancillary data layers to assist in the modeling of land cover types. We have completed a draft land cover map of central Utah in order to formalize image processing and interpretation standards and are applying these results to a second mapping region. Results from the initial land cover map of the San Rafael Swell region of Utah shows that land cover mapping at the Alliance level will be problematic, hampered by the requirement of significantly larger numbers of field training sites and the inability of imagery and ancillary data to discern between spectrally and ecologically similar alliances. This difficulty was anticipated early in the project-planning phase, and it was understood that similar alliances would need to be merged for effective mapping. To mitigate this, NatureServe and The Nature Conservancy are developing the Ecological Systems classification scheme to better match the scale of mapping provided by Landsat Enhanced Thematic Mapper imagery and available ancillary data. A preliminary test with this new classification system has shown improved results with little loss in ecological significance of the final map product.

A FUZZY ACCURACY ASSESSMENT OF THE MID-ATLANTIC GAP ANALYSIS LAND COVER MAP USING AIRBORNE VIDEOGRAPHY

Rasberry, D. Ann1, and Alexa J. McKerrow2
1Maryland Gap Analysis Project, Princess Anne, MD; darasberry@mail.umes.edu
2North Carolina Gap Analysis Project, Raleigh, NC; mckerrow@unity.ncsu.edu

An accuracy assessment has been done for the final Mid-Atlantic Gap Analysis land cover maps. The 62 land cover types mapped in the three-state area (Maryland, Delaware, New Jersey) represent types that vary in their ecological and structural similarity, therefore a fuzzy assessment was used to determine the accuracy of the maps. A stratified random sampling design with a target of forty assessment points per cover class was used. A systematic grid of airborne videography data was used to gather the assessment points. The final accuracies are weighted based on the areal extent of the mapped classes. The fuzzy error matrix used in the assessment and the per-class accuracies will be presented. The patterns of land cover and errors in the Mid-Atlantic study area will be discussed with respect to the suitability of the map for a variety of uses.

MAPPING PLANT COMMUNITIES IN THE MOJAVE DESERT OF SOUTHERN NEVADA USING CLASSIFICATION TREES: A PRELIMINARY ASSESSMENT OF MAPPING NVCS ALLIANCES

Sajwaj, Todd D.1, William G. Kepner 2, and David F. Bradford 2
1U.S. Army Corps of Engineers, co-located at U.S. EPA, 944 E. Harmon Ave., Las Vegas, NV 89109; (702) 798-2279; Sajwaj.Todd @epa.gov
2U.S. Environmental Protection Agency, 944 E. Harmon Ave., Las Vegas, NV 89019

In our efforts to map the vegetation of the Mojave Desert of southern Nevada, we developed a methodology to map plant communities via classification and regression trees, or CARTs. The target for the Southwest Regional GAP (SW ReGAP) project's vegetation classification efforts is the National Vegetation Classification System's (NVCS) Alliance Level, a goal supported by the National Gap Analysis Program (USGS BRD) and Federal Geographic Data Committee (FGDC). It is unclear whether this goal is realistic given constraints of time and money on field data collection. The field data used to construct land cover models consists of two essential types: 1) characterization of the plant community at a site, and 2) GIS polygons delineating the site. Plant community data were used to assign an NVCS alliance name to the site polygon. Site polygons were intersected with Landsat ETM+ imagery, DEMs, and various GIS data layers, resulting in a data set in which the alliance name was associated with approximately 60 spectral, topographic, climatic, and edaphic characteristics for each of 2000+ sites. The site polygon data set was then subdivided into coarse elevation and landform strata to reduce its complexity for modeling purposes. The vegetation alliances within each stratum were modeled separately and then re-intregrated to produce the final model for the Mojave region. The final model produced a vegetation classification that was assessed for thematic accuracy. Finally, the number of alliances differentiated by the CART model was compared with the number of listed NVCS alliances for the Mojave region as a preliminary assessment of the feasibility of alliance-level mapping.

A NEW APPROACH FOR TESTING THE ACCURACY OF VERTEBRATE OCCURRENCE PREDICTIONS FROM GAP ANALYSES

Schaefer, Sandra M.1, William B. Krohn2, and Raymond J. O'Connor1
1Department of Wildlife Ecology, University of Maine, Orono, Maine 04469-5755
2USGS Biological Resources Division, Maine Cooperative Fish and Wildlife Research Unit, University of Maine, Orono, Maine 04469-5755

Traditionally, the accuracy of models that predict the occurrences of multiple species of wildlife based on habitat (i.e., habitat relations models), as in Gap Analysis, are tested using field data collected from test locations within a study area. Commission and omission errors are calculated on a site-by-site basis by comparing model predictions to field data across taxa at the site (i.e., site-specific approach). In the present study, we use avian predictions from the Maine Gap Analysis to propose a complementary approach that improves on the traditional point estimates of error by computing bounded estimates of the error rates by species across sites-a species-specific approach. Two alternative assumptions concerning the completeness of field survey data were used to provide upper and lower bounds for the error rates calculated by the species-specific approach. First, field surveys for all species were assumed to be complete (i.e., all species using the site were inventoried). The contrasting assumption was that field data were incomplete, meaning species seemingly overpredicted by the models were actually missed by the field inventories. With these two assumptions it was possible to bracket the true situation by calculating an Error Range of commission for each species (we had too few species missed by the models [n = 6] to calculate this statistic for omission errors). The error range for commission, and therefore the uncertainty as to how accurately the species was actually predicted, was smaller for widespread species than for more restricted species (rho = -0.58, P < 0.01) and for species easily detected in the field than for more difficult species (rho = -0.66, P < 0.01). If a species has a high error range for commission and a high likelihood of being detected in a field survey, then an overprediction is probably true, and the model is in error. In contrast, if a species has a high commission error range and low likelihood of being detected, then an error of overprediction is probably not a real error, but instead the result of incomplete field surveys for that species. The same reasoning can be used for those species reporting low commission error ranges. A species with a low commission error range and a low likelihood of detection has probably been well modeled, whereas a species with a low commission error range and a high detectability suggests that the model or the model process may be causing the overpredictions. Because the species-specific approach provides insight into what species are being modeled poorly, while the site-specific approach indicates how well the species in various geographic areas are modeled, we recommend using both approaches when testing wildlife-habitat models.

HUDSON RIVER VALLEY HABITAT VULNERABILITY ASSESSMENT PROJECT

Smith, Stephen D., Warren A. Brown, and Charles Smith Cornell Institute for Resource Information Systems, Cornell Institute for Social and Economic Research, and College of Agriculture and Life Sciences, Department of Natural Resources, Center for the Environment 302 Rice Hall, Ithaca, NY 14853; (607) 255-0803; sds3@cornell.edu

The goal of this project is to develop a CENSUS-based methodology which uses projected population growth information, census data, and physiographic restrictions to assign a habitat vulnerability index to individual census block groups. This habitat vulnerability index could then be used as a guide in land use policy decisions. The Hudson River Valley (HRV) provides habitat for hundreds of migratory and resident species of wildlife. Recent data developed by the New York Gap Analysis (NY-GAP) project shows that over 80% of the terrestrial vertebrate species within NYS can be found in the HRV. Using fine-filter and coarse-filter categories, the NY-GAP project has identified how the HRV contributes to statewide diversity of vegetation associations and terrestrial vertebrates. In general, the variety of herpetofauna found in the HRV, especially turtles, allows the HRV to offer unique conservation opportunities not found elsewhere in New York State. The HRV also plays an equally important role in the state's cultural heritage and economic stability. It supports over half of New York State's human population and is under constant development pressure. The HRV is facing a period of re-industrialization and concomitant residential development. This re-industrialization and residential growth will impact the biodiversity of the region and thereby the ecology of the Hudson River. There are few things that have as great an impact on the ecology of a place as does human development. The ability to predict accurately the loci of that development and subsequently identify the ecologically and culturally sensitive areas susceptible to the resulting impacts will empower decision makers with the knowledge necessary to take actions required to minimize impacts and maintain wildlife and fish habitat, biological diversity, and regionally significant historical/cultural sites.

GAP ANALYSIS OF SPECIAL FEATURES: GENERAL DISCUSSION AND A CASE STUDY OF AMPHIBIANS AND MAMMALS OF WEST VIRGINIA'S CAVES

Strager, Jacquelyn M.1, and Charles B. Yuill2
1Natural Resource Analysis Center, P.O. Box 6108, West Virginia University, Morgantown, WV 26506-6108; (304) 293-4832 ext.4455; jrowe@wvu.edu
2Natural Resource Analysis Center, P.O. Box 6108, West Virginia University, Morgantown, WV 26506-6108; (304) 293-4832; cyuill@wvu.edu

As part of the overall Gap Analysis mission to compile information in support of biodiversity maintenance, individual GAP projects may conduct analyses focusing on special features of states or regions. Special feature analysis may be used to capture location information for specialized or localized habitat features such as caves, springs, cliffs, certain tree species, etc. These special features may not normally be found in statewide land cover and habitat data sets but may be important factors in the distribution of certain wildlife species. In addition, the assignment of GAP management status (level of protection) may also be specifically tailored to the special features analyzed. A case study describing West Virginia Gap Analysis (WV-GAP) work on amphibians and mammals utilizing cave habitats will be presented. In addition to standard Gap Analysis methods, WV-GAP used existing data sets for cave locations and cave accessibility to determine the gap analysis status of cave species in the state. Additional examples of special features of interest will also be discussed.

ASSESSING ACCURACY ASSESSMENT IN GAP ANALYSIS: EXAMINING WHEN IT MAY BE NEITHER ACCURACY NOR ASSESSMENT

Thompson, Bruce C.1, Kenneth G. Boykin2, and Robert A. Deitner2
1USGS-BRD, New Mexico Cooperative Fish and Wildlife Research Unit, New Mexico State University, P.O. Box 30003 MSC 4901, Las Cruces, NM 88003; (505) 646-6093; bthompso@nmsu.edu
2New Mexico Cooperative Fish and Wildlife Research Unit and Fishery and Wildlife Sciences Department, New Mexico State University, P.O. Box 30003 MSC 4901, Las Cruces, NM 88003; (KGB) (505) 646-6303, kboykin@nmsu.edu; (RAD) (505) 646-3355, bdeitner@nmsu.edu

The national Gap Analysis Program historically has incorporated accuracy assessments regarding land cover mapping and animal predictive modeling as part of project expectations. These expectations have been met in various ways among final products, although assessments sometimes have not been performed. Assessments of land cover maps have struggled with questions of sampling sufficiency and data source(s) to use. Animal prediction assessments have typically engaged some version of cross-checking "listings" of animals associated with specific areas against species predictions for those areas. These approaches can and have missed the mark with regard to what is assessed and whether an interpretable statement of accuracy can be made. A key element in designing an accuracy assessment is that independent data are used to examine a predicted metric in comparison to measured or observed evidence of the same metric. In either case, design requires that the prediction be a product of a committed (unchanging) classification for which an accuracy statement will be prepared and provided to users. For land cover, this premise requires that a committed land cover predictive map based on a specified classification and resolution be examined with data that are indicative of scale and time frame applicable to the landscape predicted. Optimally, this should involve data collected independent of map development and translated to the resolution of the predictive map. These features are standards to judge such assessment against. For predicted distribution of animal habitat (which gap analysis actually predicts), the premise requires that such predictions be assessed with data about presence of appropriate habitat features at sample locations, not whether there is a record of occurrence for the species. As such, traditional assessment of animal (habitat) prediction maps in gap analysis, using checklist data, are fundamentally flawed by not examining the appropriate metric mapped and possibly not correctly scaled to gap analysis predictions. Gap analysis researchers and project administrators are cautioned to re-examine merits of assessments performed to date and, more importantly, to more critically question future accuracy assessment designs and proposals.

GAP MAPPING OF PREDICTED SPECIES DISTRIBUTIONS: PERSPECTIVES ON ISSUES OF STATISTICAL INFERENCE

Umphrey, Gary Department of Mathematics and Statistics, University Of Guelph, Guelph, Ontario, Canada N1G 2W1; umphrey@uoguelph.ca

A tremendous surge of interest in species distribution modeling has been substantially motivated by the importance of predicted species distributions in gap analysis and other geospatially explicit biodiversity assessment methodologies. Typical GAP modeling attempts to capture "all available information" from multiple and diverse information sources, such as studies of habitat-species affinities, point locality records, and expert opinion. However, both the complexity of the phenomenon being modeled and the complexity of available information generate many questions on how best to model and, in view of the many known problems in data quantity and quality, even the scientific validity of any modeling process. I propose that, following a suggestion of Arthur Dempster, Ronald Fisher's concept of Mathematical Probability provides a scientifically sound inferential model for interpreting and refining GAP models of predicted species distributions. I also discuss how desirable statistical properties for species maps provide guidance for prioritizing the allocation of effort and resources for acquiring new information.

CHALLENGES OF MAPPING THE LAND COVER OF ILLINOIS

Weicherding, Tari A., Brooke A. Bahnsen, Leonardo Chapa, and Patrick W. Brown Illinois Natural History Survey, Center for Wildlife Ecology, 607 E. Peabody Drive, Champaign, IL 61820; tweicher@uiuc.edu; bahnsen@uiuc.edu; l-chapa1@uiuc.edu; pbrown@mail.inhs.uiuc.edu

Development of an accurate land cover classification layer for GAP has been a difficult challenge in this agricultural landscape with highly fragmented forested areas and small isolated wetlands. The Illinois Department of Natural Resources (IDNR), the U.S. Department of Agriculture National Agricultural Statistics Service (USDA-NASS), and the Illinois Department of Agriculture (IDA) in 1999 formally established a Geo-Spatial Information System Partnership for the purpose of producing statewide land use and land cover information on a recurring basis. Remote sensing analysts from this multiagency partnership are working on an Illinois Interagency Landscape Classification Project to classify the Illinois landscape using Landsat Thematic Mapper satellite imagery from 1999 and 2000. This Illinois Interagency Landscape Classification data set will be incorporated into the final GAP land cover classification layer. The increased level of detail of the new land cover classification, especially for agricultural areas, provides useful information to scientists and planners. Details about crop composition, watersheds, and changes in the urban fringe provide information planners' need to ensure sustainable development. Based upon this partnership, we determined that the Illinois landscape as a whole is 7.67% urban/developed, 63.62% agricultural lands, 11.01% grassland, <1.00% nonforested wetlands, 1.68% water, and 14.83% forest/woodland. Most of the forested lands are highly fragmented except the Shawnee National Forest, which occurs only in extreme southern Illinois. Concerns over accuracy assessment structured our approach. The land cover was manually edited by someone other than the person who classified the data. The editing helped correct misclassification or mixing of spectrally similar ground features. Next, the land cover was accuracy assessed, by another unbiased person, using a stratified random sampling method and using aerial photos. Based upon our experience in Illinois, we believe this partnership has improved the accuracy of our land cover map, and a similar partnership like this one would be beneficial to other states that are dominated by agriculture.

USING GAP ANALYSIS DATA IN AN ASSESSMENT OF THE POTENTIAL IMPACTS OF FUTURE MOUNTAINTOP MINING IN WEST VIRGINIA

Yuill, Charles, and Jackie Strager West Virginia University, Natural Resource Analysis Center - The Davis College, 1138 Agricultural Sciences Building, Morgantown, WV 26506-6108; (304) 293-4832; cyuill@wvu.edu

One of the first critical uses of the West Virginia GAP data was in an Appalachian-wide environmental assessment of the impacts of mountaintop coal mining. This environmental assessment was initiated in 1999 as a required multi-federal agency response to court actions that were undertaken by a consortium of environmental organizations that sued over state implementation of the Federal Surface Mining and Reclamation Control Act (SMCRA). These actions were based on potential violations of the Clean Water Act by various state's mining regulations. The environmental assessment that is being completed is comprehensive, focusing on aspects ranging from socioeconomic and community impacts to hydrologic and wildlife impacts. Areas that are potentially suitable for mountaintop mining were identified in associated GIS-based modeling efforts. These data then provided the base for estimating the impacts of potential future mining scenarios on land cover, land use, community development, and wildlife-assessments that relied primarily on the GAP-generated land cover and wildlife models that were developed. The assessment also allowed for the first revision of the GAP land cover map, though these revisions focused primarily on developed and disturbed land cover types.

USGS ||Gap Analysis Program || NBII ||
Disclaimer || Privacy || Accessibility||
GAP Webmaster ||

Presentations for 2002 Meeting

The following list represents all of the presentations at the Shepherdstown Meeting. Click on a title to read the abstract. Click on the abstract title to get to the powerpoint presentation.

Keynote Address

Land Cover

Aquatic GAP

Applications and Socioeconomics

Animal Modeling

Accuracy Assessment

PROSPECTS AND CHALLENGES FOR GAP ANALYSIS IN AQUATIC CONSERVATION

BIOGEOGRAPHY, CONSERVATION BIOLOGY, AND THE WALLACEAN SHORTFALL

Schaefer, Sandra M.1, William B. Krohn2, and Raymond J. O'Connor1 1Department of Wildlife Ecology, University of Maine, Orono, Maine 04469-5755 2USGS Biological Resources Division, Maine Cooperative Fish and Wildlife Research Unit, University of Maine, Orono, Maine 04469-5755

Umphrey, Gary Department of Mathematics and Statistics, University Of Guelph, Guelph, Ontario, Canada N1G 2W1; umphrey@uoguelph.ca

Weicherding, Tari A., Brooke A. Bahnsen, Leonardo Chapa, and Patrick W. Brown Illinois Natural History Survey, Center for Wildlife Ecology, 607 E. Peabody Drive, Champaign, IL 61820; tweicher@uiuc.edu; bahnsen@uiuc.edu; l-chapa1@uiuc.edu; pbrown@mail.inhs.uiuc.edu

Yuill, Charles, and Jackie Strager West Virginia University, Natural Resource Analysis Center - The Davis College, 1138 Agricultural Sciences Building, Morgantown, WV 26506-6108; (304) 293-4832; cyuill@wvu.edu

Schaefer, Sandra M.1, William B. Krohn2, and Raymond J. O'Connor1
1Department of Wildlife Ecology, University of Maine, Orono, Maine 04469-5755
2USGS Biological Resources Division, Maine Cooperative Fish and Wildlife Research Unit, University of Maine, Orono, Maine 04469-5755