Gap analysis of animal species distributions in Maryland, Delaware, and New Jersey

              A GAP ANALYSIS OF
Animal Species Distributions in
MARYLAND, DELAWARE, AND
NEW JERSEY
2006 Final Report
A GEOGRAPHIC APPROACH TO PLANNING FOR BIOLOGICAL DIVERSITY
U.S. Department of the Interior
U.S. Geological Survey
ii
THE MARYLAND, DELAWARE, AND
NEW JERSEY
GAP ANALYSIS PROJECT
FINAL REPORT – Part 2: Vertebrate Species Distributions
December 27, 2006
Principal and Co-Principal Investigators:
Richard C. McCorkle, U.S. Fish & Wildlife Service
James N. Gorham, U.S. Fish & Wildlife Service
D. Ann Rasberry, University of Maryland Eastern Shore
Research Associates:
Thomas F. Breden, Rebecca A. Eanes, Paula G. Becker, Dana L. Limpert
Contract Administration Through:
U.S. Fish & Wildlife Service, Delaware Bay Estuary Project
University of Maryland Eastern Shore
Submitted by:
Richard C. McCorkle
Research Performed Under:
Interagency Agreement No. 14-45-0009-94-990
Cooperative Agreement Nos. 14-48-50181-99-J-006, 14-48-0005-93-9061
U.S. Fish & Wildlife Service
Delaware Bay Estuary Project
U.S. Geological Survey
Biological Resources Division
Gap Analysis Program
iii
Suggested Citation:
McCorkle, R.C., J.N. Gorham, and D.A. Rasberry. 2006. Gap Analysis of Animal
Species Distributions in Maryland, Delaware, and New Jersey. Final Report – Part 2.
U.S. Fish & Wildlife Service, Delaware Bay Estuary Project, and USGS Biological
Resources Division, Gap Analysis Program. 229 pp.
iv
Table of Contents
List of Tables ................................................................................................................... viii
List of Figures .................................................................................................................... ix
Executive Summary........................................................................................................... xi
Acknowledgements........................................................................................................... xv
Chapter 1 – Introduction ......................................................................................................1
1.1 How This Report is Organized..................................................................................1
1.2 The Gap Analysis Program Mission .........................................................................1
1.3 The Gap Analysis Concept........................................................................................2
1.4 General Limitations...................................................................................................4
1.5 The Study Area..........................................................................................................4
Chapter 2 – Predicted Animal Species Distributions and Species Richness .......................8
2.1 Introduction ...............................................................................................................8
2.2 Methods.....................................................................................................................9
2.2.1 Mapping Standards and Data Sources...............................................................9
2.2.2 Mapping Range Extent ....................................................................................11
2.2.3 Habitat Modeling Grids...................................................................................15
2.2.3.1 Habitat Types...........................................................................................15
2.2.3.2 Wetland Buffers.......................................................................................22
2.2.3.3 Forest Fragmentation Variables...............................................................24
2.2.3.4 Open – Grassaland Area ..........................................................................32
2.2.3.5 Open – Edge Habitat................................................................................32
2.2.3.6 Land Form (Elevation, Slope, Aspect) ....................................................32
2.2.3.7 Road Juxtaposition ..................................................................................32
2.2.3.8 Forest Juxtaposition.................................................................................33
2.2.3.9 Special Habitat Features ..........................................................................35
2.2.3.9.1 Island................................................................................................35
2.2.3.9.2 Cave .................................................................................................35
2.2.3.9.3 Outcrop ............................................................................................36
2.2.3.9.4 Cliff..................................................................................................37
2.2.3.9.5 Dam/Bridge......................................................................................37
2.2.3.9.6 Combining Special Habitat Features ...............................................37
2.2.4 Wildlife Habitat Relationships ........................................................................38
2.2.4.1 MDN-GAP Species List ..........................................................................38
2.2.4.2 Development of Wildlife Habitat Relationships Models ........................39
2.2.5 Distribution Modeling.....................................................................................41
v
2.3 Results .....................................................................................................................42
2.3.1 Birds ................................................................................................................42
2.3.2 Mammals .........................................................................................................44
2.3.3 Reptiles............................................................................................................44
2.3.4 Amphibians .....................................................................................................47
2.4 Species Richness .....................................................................................................49
2.4.1 Bird Species Richness .....................................................................................49
2.4.2 Rare Bird Species Richness.............................................................................49
2.4.3 Mammal Species Richness..............................................................................53
2.4.4 Rare Mammal Species Richness .....................................................................53
2.4.5 Reptile Species Richness.................................................................................56
2.4.6 Rare Reptile Species Richness ........................................................................56
2.4.7 Amphibian Species Richness ..........................................................................56
2.4.8 Rare Amphibian Species Richness..................................................................60
2.4.9 Vertebrate Species Richness – All Taxonomic Groups ..................................60
2.4.10 Rare Vertebrate Species Richness .................................................................60
2.5 Accuracy Assessment..............................................................................................65
2.5.1 Methods ...........................................................................................................65
2.5.2 Results .............................................................................................................66
2.6 Limitations and Discussion.....................................................................................69
2.6.1 Species Richness .............................................................................................69
2.6.2 Vertebrate Species Distribution Model Accuracy...........................................71
2.6.3 Accuracy Assessment of Predicted Vertebrate Species Distributions.............72
Chapter 3 – Analysis Based On Stewardship and Management Status .............................74
3.1 Introduction .............................................................................................................74
3.2 Methods...................................................................................................................75
3.3 Results .....................................................................................................................76
3.3.1 Species with Less than 1% of Predicted Distribution in Status 1 or 2 ............77
3.3.1.1 Amphibians..............................................................................................77
3.3.1.2 Birds ........................................................................................................77
3.3.1.3 Mammals .................................................................................................78
3.3.1.4 Reptiles ....................................................................................................78
3.3.2 Species with Less than 10% of Predicted Distribution in Status 1 or 2 ..........78
3.3.2.1 Amphibians..............................................................................................78
3.3.2.2 Birds ........................................................................................................79
3.3.2.3 Mammals .................................................................................................80
3.3.2.4 Reptiles ....................................................................................................80
3.3.3 Species with Less than 20% of Predicted Distribution in Status 1 or 2 ..........81
3.3.3.1 Amphibians..............................................................................................81
3.3.3.2 Birds ........................................................................................................81
3.3.3.3 Mammals .................................................................................................81
3.3.3.4 Reptiles ....................................................................................................82
3.3.4 Species with Less than 50% of Predicted Distribution in Status 1 or 2 ..........82
vi
3.3.4.1 Amphibians..............................................................................................82
3.3.4.2 Birds ........................................................................................................82
3.3.4.3 Mammals .................................................................................................82
3.3.4.4 Reptiles ....................................................................................................82
3.3.5 Species with More than 50% of Predicted Distribution in Status 1 or 2.........83
3.3.5.1 Amphibians..............................................................................................83
3.3.5.2 Birds ........................................................................................................83
3.3.5.3 Mammals .................................................................................................83
3.3.5.4 Reptiles ....................................................................................................83
3.3.6 Analysis of Important Projectwide Species Assemblages...............................83
3.3.6.1 Vernal Pool-Breeding Amphibians .........................................................83
3.3.6.2 Wading Birds of Pea Patch Island ...........................................................84
3.3.7 Analysis of State Endemics .............................................................................84
3.4 Limitations and Discussion.....................................................................................84
Chapter 4 – Stewardship Status of Predicted Rare Species Richness Hotspots.................86
4.1 Introduction .............................................................................................................86
4.2 Methods...................................................................................................................86
4.3 Results .....................................................................................................................87
4.3.1 Predicted Gaps in Protection of Rare Bird Species Hotspots..........................87
4.3.2 Predicted Gaps in Protection of Rare Mammal Species Hotspots ..................87
4.3.3 Predicted Gaps in Protection of Rare Reptile Species Hotspots .....................90
4.3.4 Predicted Gaps in Protection of Rare Amphibian Species Hotspots...............90
4.3.5 Predicted Gaps in Protection of Rare Vertebrate Species Hotspots ................93
4.4 Limitations and Discussion.....................................................................................95
Chapter 5 – Conclusions and Management Implications...................................................96
Chapter 6 – Product Use and Availability .........................................................................99
6.1 How to Obtain the Products ....................................................................................99
6.1.1 Minimum GIS Required for Data Use.............................................................99
6.2 Disclaimer ...............................................................................................................99
6.3 Metadata................................................................................................................100
6.4 Appropriate and Inappropriate Uses of the Data...................................................101
Literature Cited ................................................................................................................104
Glossary of Terms ............................................................................................................111
Glossary of Acronyms......................................................................................................114
Appendix A: Examples of GAP Applications .................................................................115
vii
Appendix B: Habitat Types of the Eastern United States ................................................119
Appendix C: Table summarizing habitats defined by other authors and proposed
habitats for MDN-GAP....................................................................................................124
Appendix D: List of Habitat Types: MDN-GAP Project.................................................129
Appendix E: MDN-GAP Habitat Type Descriptions ......................................................131
Appendix F: Primary References for Compiling Habitat Requirements Information .....167
Appendix G: Habitat Requirements Data Summary Form..............................................172
Appendix H: Rare Species of the MDN-GAP Project Area ............................................174
Appendix I: Accuracy of Individual Species Models by Management Area, Based on
Comparison with Checklists ............................................................................................182
Appendix J: Gap Analysis of Vertebrate Species by Stewardship Area..........................208
Appendix K: Predicted Rare Species Hotspots on Status 3 and 4 Lands ........................221
viii
List of Tables
Table 2.1 Grids Used in Habitat Modeling........................................................................10
Table 2.2 Codes Indicating Level of Certainty of Species Breeding Occurrence in
Hexagon (Hernandez 2002) ..........................................................................................12
Table 2.3 Modeling Parameters and Suitability Thresholds for Area Sensitive Species...26
Table 2.4 System for Ranking Salamander Non-Breeding Habitat ...................................33
Table 2.5 Variables Used in Evaluating Suitability of Caves for Bat Use.........................36
Table 2.6 Database Tables Used in Modeling Species Habitat Relationships and
Distributions..................................................................................................................40
Table 2.7 Accuracy Assessment by Management Area .....................................................67
Table 3.1 Proportion of Each Taxonomic Group with 0-1%, 1-10%, 10-20%, 20-50%,
and > 50% of their Predicted Distributions in GAP Status 1 and 2 Lands ...................77
ix
List of Figures
Figure 1.1 Maryland-Delaware-New Jersey Gap Analysis Project Study Area...................6
Figure 1.2 Physiographic Provinces of the MDN-GAP Study Area....................................7
Figure 2.1 Example of a Species’ Range by Hexagon.......................................................13
Figure 2.2 Example of a Species’ Range by 7.5-Minute Quadrangle................................14
Figure 2.3 Habitat Types in New Jersey ............................................................................21
Figure 2.4 Forest Fragmentation Metrics used in Habitat Modeling.................................25
Figure 2.5 Example of Probability Curve (Robbins et al. 1989) .......................................26
Figure 2.6 Correlation of Zonal Thickness and Natural Log of Forest Area as
Determined by Robbins et al. (1989) ............................................................................28
Figure 2.7 Map Depicting Forest Area Metric...................................................................30
Figure 2.8 Forest Patch Isolation in Delaware ...................................................................31
Figure 2.9 Example of a Bird Species Distribution Map...................................................43
Figure 2.10 Example of a Mammal Species Distribution Map..........................................45
Figure 2.11 Example of a Reptile Species Distribution Map ............................................46
Figure 2.12 Example of an Amphibian Species Distribution Map....................................48
Figure 2.13 Predicted Bird Species Richness for the MDN-GAP Study Area ..................50
Figure 2.14 Predicted Rare Bird Species Richness for the MDN-GAP Study Area..........51
Figure 2.15 Predicted Rare Bird Species Hotspots in the MDN-GAP Study Area ...........52
Figure 2.16 Predicted Mammal Species Richness for the MDN-GAP Study Area...........54
Figure 2.17 Predicted Rare Mammal Species Richness for the MDN-GAP Study Area ..55
Figure 2.18 Predicted Reptile Species Richness for the MDN-GAP Study Area..............57
Figure 2.19 Predicted Rare Reptile Species Richness for the MDN-GAP Study Area .....58
x
Figure 2.20 Predicted Amphibian Species Richness for the MDN-GAP Study Area .......59
Figure 2.21 Predicted Rare Amphibian Species Richness for the MDN-GAP Study
Area ...............................................................................................................................61
Figure 2.22 Predicted Vertebrate Species Richness for the MDN-GAP Study Area.........62
Figure 2.23 Predicted Rare Vertebrate Species Richness for the MDN-GAP Study
Area ...............................................................................................................................63
Figure 2.24 Management Areas Included in MDN-GAP Vertebrate Model
Accuracy Assessment....................................................................................................68
Figure 4.1 Predicted Rare Bird Species Hotspots on Status 4 Lands.................................88
Figure 4.2 Predicted Rare Mammal Species Hotspots on Status 4 Lands .........................89
Figure 4.3 Predicted Rare Reptile Species Hotspots on Status 4 Lands ............................91
Figure 4.4 Predicted Rare Amphibian Species Hotspots on Status 4 Lands......................92
Figure 4.5 Predicted Rare Vertebrate Species Hotspots on Status 4 Lands.......................94
xi
Executive Summary
Gap analysis provides an overview of the distribution and conservation status of several
components of biodiversity. There are five major objectives of the national Gap Analysis
Program: (1) map actual vegetation as closely as possible to the Alliance level; (2) map
predicted distributions of animals for which adequate distributional records, habitat
associations, and mapped habitat variables are available; (3) document occurrence of
vegetation types that are inadequately represented (gaps) in special management areas; (4)
document occurrence of animal species that are inadequately represented (gaps) in special
management areas; and (5) make all information available to resource managers and land
stewards in a readily accessible format.
To meet national objectives, gap analysis is conducted at the state level while maintaining
consistency with national standards. The Maryland-Delaware-New Jersey Gap Analysis
Project (MDN-GAP) involved the efforts of researchers from various government natural
resource agencies and universities in all three states, with the bulk of the work and project
administration being carried out by the U.S. Fish & Wildlife Service, Maryland
Department of Natural Resources, University of Maryland Eastern Shore Cooperative
Fish & Wildlife Research Unit, and New Jersey Department of Environmental Protection.
The three-state project area includes a complex mixture of habitats, ranging from coastal
beaches and estuarine tidal marshes to montane forests and bogs, and human-dominated
urban and agricultural landscapes. Despite the high degree of human land use pressure
and habitat fragmentation in many parts of the project area, there remain many
exceptional examples of regionally and globally significant natural features and wildlife
populations.
This report pertains only to the mapping and assessment of animal species distributions,
and is a supplement to an earlier report describing the development and assessment of the
vegetation and land stewardship components of this project. Animal species habitat
modeling and distribution mapping involved the development of three primary data sets:
(1) breeding ranges for all animal species; (2) a species-habitat association database with
tables that identify relationships between animal species and various habitat variables;
and (3) geographic information system (GIS) thematic layers representing the habitat
variables for which habitat relationships have been recorded in the database tables.
The ranges or distributional limits of animal species were developed primarily through
the Biodiversity Research Consortium (BRC), now administered by Nature Serve. The
BRC uses the hexagons utilized by EPA’s Environmental Monitoring and Assessment
Program. Within the Maryland-Delaware-New Jersey project area, these hexagons range
in size from about 648 to 651 square kilometers per hexagon. Each hexagon was
assigned a code reflecting the level of certainty associated with the species occurrence
data. In general, hexagons with “confirmed” or “probable” occurrence records were
included in a species’ range. For rare, threatened, or endangered species in Maryland and
Delaware, 7.5-minute quadrangles, which are significantly smaller than hexagons, were
xii
used to map ranges in order to avoid over-estimating the distributions of these rare
species. Rare species data were not available for most of New Jersey, but Breeding Bird
Atlas data were used to populate quad records for rare bird species in this state.
Development of the wildlife habitat relationships database began with a review of the
literature and compilation of habitat requirements information into an individual
summary document for each species. This document was then used as a reference in
filling out a standard data form where habitat types and other variables (e.g., elevation)
were assigned suitability rankings and relative weightings (i.e., relative influence on
species preferred habitat and geographic distribution). These habitat suitability rankings
and habitat variable weightings were then entered into tables in the wildlife habitat
relationships database. The list of habitats was developed through a review of several
other efforts to define wildlife habitats, and by identifying the particular habitat types that
are commonly mentioned in the literature.
The habitat type map was developed from three primary data sources: (1) MDN-GAP
Land Cover data; (2) National Wetlands Inventory data; and (3) National Land Cover
Data. Other habitat variables used in modeling animal species’ distributions included
proximity to wetlands (14 wetland types; 4 buffer distances), forest interior, forest patch
isolation, riparian forest width, grassland area, edge habitat, elevation, slope, aspect,
juxtaposition to forest, juxtaposition to roads, and proximity to a special habitat feature
(e.g., island, cave, outcrop, cliff, bridge).
Predictive habitat models and distribution maps were developed for 363 animal species
(206 bird species, 69 mammal species, 47 reptile species, 41 amphibian species). Bird
habitat models and distribution maps were limited to those species that regularly nest
within the project area. Although there are regionally and globally significant migratory
bird staging areas in Maryland, Delaware and New Jersey, project resource limitations
prevented inclusion of species that use the area during migration but do not nest here.
Also, there are currently many complementary efforts that are focused on addressing the
needs of these migratory bird concentrations. In addition to mapping predicted
distributions of individual species, analyses were conducted in order to identify and map
species-rich areas or “hotspots.” These analyses resulted in the identification of bird
species hotspots, mammal hotspots, reptile hotspots, and amphibian hotspots. In
addition, rare species hotspots were identified for each of these groups, and for all groups
combined.
An accuracy assessment was undertaken, comparing predicted animal distributions with
documented occurrences in managed areas (e.g., National Wildlife Refuges). The goal of
GAP is to produce maps that predict animal species distributions with an accuracy of
80% or higher. A total of 12 managed areas had species checklists to which predicted
distributions were compared. Of the 363 species modeled, 280 (77.1%) were included on
at least one of the checklists. For birds, matches between checklists and modeled
distributions exceeded 80% in only 5 of 12 areas, but exceeded 79% in 9 of these areas.
Many of the non-matches were actually caused by errors in checklists. For example,
xiii
disagreements between Breeding Bird Atlas data and checklists often corresponded with
recorded “errors.” For mammals, matches exceeded 80% in only 1 of 3 areas for which
mammal checklists existed. For reptiles, matches exceeded 80% in 3 of 4 areas, with the
lowest rate of agreement being 78.8%. For amphibians, matches exceeded 80% in only 1
of 4 areas, but significant errors were found in the checklist for at least one of the
management areas included in this comparison. Also, some checklists indicated a lack of
certainty regarding the presence of certain secretive species, and many checklists
indicated that the species included were known to occur on or “near” the management
area. A more thorough accuracy assessment, including additional expert review, is
needed to better determine the level of accuracy of animal species habitat models and
distribution maps.
The final step of gap analysis involves intersecting the distributions of elements of
biological diversity (i.e., land cover types and animal species) with the land stewardship
and management status map, in order to identify “gaps” in protection. The land
stewardship data set includes land ownership boundaries and land stewardship status
rankings that reflect the degree to which each area is managed for biodiversity, with status
1 lands affording the highest level of protection and status 4 lands providing the least
amount of protection. The predicted distributions of all 363 animal species were
intersected with the land stewardship map to produce summaries of protection for each
species. Birds and reptiles appear to have the best representation within status 1 and 2
lands, with over 15% of bird species and over 10% of reptile species having more than
10% of their potential habitat receiving these higher levels of protection. Amphibians
appear to have received the least amount of protection, with over 95% of amphibian
species having less than 10% of their potential habitat occurring within status 1 and 2
lands. When considering native species only, nearly 97% of mammal species and over
88% of all species have less than 10% of their predicted distributions occurring within
status 1 or 2 lands. Overall, it appears that all groups are poorly represented within GAP
status 1 and 2 lands.
In general, the habitats supporting the species of greatest conservation concern (i.e., those
that are rare to extremely rare within the project area and are underrepresented in status 1
and 2 lands) include early successional habitats, unpolluted mountain streams, vernal
pools (non-tidal, isolated, seasonally flooded wetlands) with substantial upland forest
buffers, forested wetlands and freshwater marshes, forest interior, broad riparian and
floodplain forests, and beach and dune habitats.
The most prominent rare species hotspots (i.e., areas with high rare species richness) that
are unprotected include the Youghiogheny River corridor and other riparian forests in
western Maryland, and some of the riparian and headwater forests of the New Jersey
Highlands and Kittatinny Mountain; forest-swamp ecotones in parts of the New Jersey
Pine Barrens; the large concentration of coastal plain ponds (i.e., vernal pools) and
surrounding hardwood forests in the Blackbird-Millington Corridor of Delaware and
Maryland; Potomac River and C&O Canal tributaries northwest of Washinton, D.C.; and
xiv
wetlands associated with headwaters and tributaries of several rivers in the southern Pine
Barrens and Highlands of New Jersey.
The results of this effort identify many species of conservation concern and habitats that
are in need of additional protection. These results should be incorporated into
conservation planning efforts and used to guide additional field investigations. Such
investigations and expert review of the results may also lead to a better understanding of
data limitations and ways of refining and improving the data.
xv
Acknowledgements
Thanks to Amos Eno and the staff of the National Fish and Wildlife Foundation, who
funded the early development of the GAP concept and to the originators including J.
Michael Scott, Blair Csuti, and Jack Estes and the pioneering scientists who forged the
way. Thanks to John Mosesso and Doyle Frederick of the U.S. Geological Survey
(USGS) Biological Resource Division (BRD) Office of Inventory and Monitoring, for
their support of the national Gap Analysis Program, especially during its transition from
the U.S. Fish and Wildlife Service to the National Biological Service and then to the U.S.
Geological Survey BRD. Thanks to Reid Goforth and the staff at the USGS BRD
Cooperative Research Units for administering Gap's research and development phase
from headquarters. Without those mentioned above, there could not have been a Gap
Analysis Program. Thanks also to the staffs of theNational Gap Analysis Program,
Center for Biological Informatics, and Biological Resources Division headquarters. We
also acknowledge contributions to this report by Chris Cogan, Patrick Crist, Blair Csuti,
Tom Edwards, Michael Jennings, and other GAP researchers.
Many thanks to David Hannah for his early contributions to this project, to Dave Stout
and Teresa Burrows for their administrative support during the early stages of the project,
and to Dave Wrazien for his early GIS support. Thanks to Ed Christoffers, Gregory
Breese, and Barbara Van Leer for their administrative support, to Mickey Hayden for IT
support throughout the project (and for helping me figure out MS Word at the end), and
to Flavia Rutkosky for her moral support. Thanks also to the many people in the USFWS
Regional Office in Hadley, Massachusetts for the administrative support they provided
throughout the project.
Thanks to Kitt Heckscher, Gene Hess, Dorothy Hughes, Pilar Hernandez, Larry Master,
Rob Solomon, Winston Wayne, Rick West, Scott Smith, Vince Elia, Bill Grogin, Roland
Roth, Jim White, Mick McLaughlin, Paul Kerlinger, and Smithsonian Institution staff at
the National Museum of Natural History for their valuable contributions to the
development of species’ range maps. Thanks to Lynn Broaddus, Karen Bennett and Lynn
Davidson for their assistance and cooperation in developing range maps for rare,
threatened, and endangered species.
We are grateful to Larrry Thornton and John Tyrawski for providing New Jersey GIS
Resource Data, to Larry Pomatto for providing wetlands data for Delaware, to Ted
Webber for reviewing and commenting on some of the early bird models, to Steve Bittner
for his early contributions to the black bear model, and to Richard DeGraaf for sharing
results of his research.
Special thanks to Chan Robbins, Cherry Keller, Deanna Dawson and John Sauer for their
assistance in developing forest fragmentation metrics and suitability thresholds for area-sensitive
forest birds, and for providing data from their research. Thanks also to Mike
Erwin and other Patuxent researchers who gave generously of their time.
xvi
Thanks to Bill McAvoy, Peter Bowman and Keith Clancy for sharing their knowledge of
plant communities, to Rob Line, Phil Carpenter, Ron Vickers and Tim Palmer for their
assistance in developing the Land Stewardship data, and to Steve Atzert, Frank Smith,
George O’Shea and other USFWS Region 5 Refuge personnel for sharing their personal
knowledge of National Wildlife Refuge lands they manage.
We’d also like to thank Jim Hall, Holliday Obrecht, Christopher Wicker, Rachael Chiche,
Connie Skipper, Walter Ellison, Sarah Milbourne, Katherine Whittemore, and Annie
Larson for providing species checklists and information pertaining to species occurrences
on government-owned lands.
1
Chapter 1: Introduction
1.1 How This Report is Organized
This report is a summation of a scientific project. While we endeavor to make it
understandable for as general an audience as practicable, it reflects the complexity of the
project it describes. A glossary of terms is provided to aid the reader in its understanding,
and for those seeking a detailed understanding of the subjects, the cited literature should
be helpful. The organization of this report follows the general chronology of project
development, beginning with the production of the individual data layers and concluding
with analysis of the data. It diverges from standard scientific reporting by embedding
results and discussion sections within individual chapters. This was done to allow the
individual data products to stand on their own as testable hypotheses and provide data
users with a concise and complete report for each data and analysis product.
This is a supplement to a previously published final report describing the land cover and
land stewardship mapping components of the project. The animal species distribution
mapping was not completed in time for inclusion in that report, and is instead presented
here. We begin this report with an overview of the Gap Analysis Program mission,
concept, and limitations. We then present a synopsis of how the current biodiversity
condition of the project area came to be, followed by animal species distribution
prediction and species richness analyses. Data development leads to the Analysis section,
which reports on the status of the elements of biodiversity (animal species) for Maryland,
Delaware and New Jersey. Finally, we describe the management implications of the
analysis results and provide information on how to acquire and use the data.
1.2 The Gap Analysis Program Mission
The mission of the Gap Analysis Program is to prevent conservation crises by providing
conservation assessments of biotic elements (plant communities and native animal
species) and to facilitate the application of this information to land management
activities. This is accomplished through the following five objectives:
1) map actual land cover as closely as possible to the alliance level (FGDC 1997).
2) map the predicted distribution of those terrestrial vertebrates and selected other taxa
that spend any important part of their life history in the project area and for which
adequate distributional habitats, associations, and mapped habitat variables are
available.
3) document the representation of natural vegetation communities and animal species in
areas managed for the long-term maintenance of biodiversity.
4) make all GAP project information available to the public and those charged with land
use research, policy, planning, and management.
5) build institutional cooperation in the application of this information to state and
regional management activities.
2
To meet these objectives, it is necessary that GAP be operated at the state or regional
level but maintain consistency with national standards. Within the state, participation by
a wide variety of cooperators is necessary and desirable to ensure understanding and
acceptance of the data and forge relationships that will lead to cooperative conservation
planning.
1.3 The Gap Analysis Concept
The Gap Analysis Program (GAP) brings together the problem-solving capabilities of
federal, state, and private scientists to tackle the difficult issues of land cover mapping,
animal habitat characterization, and biodiversity conservation assessment at the state,
regional, and national levels. The program seeks to facilitate cooperative development
and use of information. Throughout this report we use the terms "GAP" to describe the
national program, "GAP Project" to refer to an individual state or regional project, and
"gap analysis" to refer to the gap analysis process or methodology.
Much of the following discussion was taken verbatim from Edwards et al. 1995, Scott et
al. 1993, and Davis et al. 1995. The gap analysis process provides an overview of the
distribution and conservation status of several components of biodiversity. It uses the
distribution of actual vegetation and predicted distribution of terrestrial vertebrates and,
when available, invertebrate taxa. Digital map overlays in a GIS are used to identify
individual species, species-rich areas, and vegetation types that are unrepresented or
underrepresented in existing management areas. It functions as a preliminary step to the
more detailed studies needed to establish actual boundaries for planning and management
of biological resources on the ground. These data and results are then made available to
the public so that institutions as well as individual landowners and managers may become
more effective stewards through more complete knowledge of the management status of
these elements of biodiversity. GAP, by focusing on higher levels of biological
organization, is likely to be both cheaper and more likely to succeed than conservation
programs focused on single species or populations (Scott et al.1993).
Biodiversity inventories can be visualized as "filters" designed to capture elements of
biodiversity at various levels of organization. The filter concept has been applied by The
Nature Conservancy, which established Natural Heritage Programs in all 50 states. The
Nature Conservancy employs a fine filter of rare species inventory and protection and a
coarse filter of community inventory and protection (Jenkins 1985, Noss 1987). It is
postulated that 85-90% of species can be protected by the coarse filter without having to
inventory or plan reserves for those species individually. A fine filter is then applied to
the remaining 15-10% of species to ensure their protection. Gap analysis is a coarse-filter
method because it can be used to quickly and cheaply assess the other 85-90% of species.
GAP is not designed to identify and aid protection of elements that are rare or of very
restricted distribution; rather it is designed to help "keep common species common" by
identifying risk far in advance of actual population decline. These concepts are further
developed below.
3
The intuitively appealing idea of conserving most biodiversity by maintaining examples
of all natural community types has never been applied, although numerous approaches to
the spatial identification of biodiversity have been described (Kirkpatrick 1983, Margules
and Nicholls 1988, Pressey and Nicholls 1989, Nicholls and Margules 1993).
Furthermore, the spatial scale at which organisms use the environment differs
tremendously among species and depends on body size, food habits, mobility, and other
factors. Hence, no coarse filter will be a complete assessment of biodiversity protection
status and needs. However, species that fall through the pores of the coarse filter, such as
narrow endemics and wide-ranging mammals, can be captured by the safety net of the
fine filter. Community-level (coarse-filter) protection is a complement to, not a substitute
for, protection of individual rare species.
Gap analysis is essentially an expanded coarse-filter approach (Noss 1987) to biodiversity
protection. The land cover types mapped in GAP serve directly as a coarse filter, the goal
being to assure adequate representation of all native vegetation community types in
biodiversity management areas. Landscapes with great vegetation diversity often are
those with high edaphic variety or topographic relief. When elevational diversity is very
great, a nearly complete spectrum of vegetation types known from a biological region
may occur within a relatively small area. Such areas provide habitat for many species,
including those that depend on multiple habitat types to meet life history needs (Diamond
1986, Noss 1987). By using landscape-sized samples (Forman and Godron 1986) as an
expanded coarse filter, gap analysis searches for and identifies biological regions where
unprotected or underrepresented vegetation types and animal species occur.
More detailed analyses were not part of this project, but are areas of research that GAP as
a national program is pursuing. For example, a second filter could combine species
distribution information to identify a set of areas in which all, or nearly all, mapped
species are represented. There is a major difference between identifying the richest areas
in a region (many of which are likely to be neighbors and share essentially the same list of
species) and identifying areas in which all species are represented. The latter task is most
efficiently accomplished by selecting areas whose species lists are most different or
complementary. Areas with different environments tend to also have the most different
species lists for a variety of taxa. As a result, a set of areas with complementary sets of
species for one higher taxon (e.g., mammals) often will also do a good job representing
most species of other higher taxa (e.g., trees, butterflies). Species with large home
ranges, such as large carnivores, or species with very local distributions may require
individual attention. Additional data layers can be used for a more holistic conservation
evaluation. These include indicators of stress or risk (e.g., human population growth,
road density, rate of habitat fragmentation, distribution of pollutants) and the locations of
habitat corridors between wildlands that allow for natural movement of wide-ranging
animals and the migration of species in response to climate change.
4
1.4 General Limitations
Limitations must be recognized so that additional studies can be implemented to
supplement GAP. The following are general project limitations; specific limitations for
the data are described in the respective sections:
1. GAP data are derived from remote sensing and modeling to make general assessments
about conservation status. Any decisions based on the data must be supported by
ground-truthing and more detailed analyses.
2. GAP is not a substitute for threatened and endangered species listing and recovery
efforts. A primary argument in favor of gap analysis is that it is proactive: it seeks to
recognize and manage sites of high biodiversity value for the long-term maintenance
of populations of native species and communities before they become critically rare.
Thus, it should help to reduce the rate at which species require listing as threatened or
endangered. Those species that are already greatly imperiled, however, still require
individual efforts to assure their recovery.
3. GAP data products and assessments represent a snapshot in time generally
representing the date of the satellite imagery. Updates are planned on a 5-10 year
cycle, but users of the data must be aware of the static nature of the products.
4. GAP is not a substitute for a thorough national biological inventory. As a response to
rapid habitat loss, gap analysis provides a quick assessment of the distribution of
vegetation and associated species before they are lost, and provides focus and
direction for local, regional, and national efforts to maintain biodiversity. The process
of improving knowledge in systematics, taxonomy, and species distributions is
lengthy and expensive. That process must be continued and expedited, however, in
order to provide the detailed information needed for a comprehensive assessment of
our nation's biodiversity. Vegetation and species distribution maps developed for
GAP can be used to make such surveys more cost-effective by stratifying sampling
areas according to expected variation in biological attributes.
1.5 The Study Area
The Maryland-Delaware-New Jersey Gap Analysis Project (MDN-GAP) study area
includes the states of Maryland, Delaware and New Jersey (Figure 1.1). Other authors
(Robbins and Blom 1996, Hess et al. 2000, Walsh et al. 1999) have described these states
in detail. In general, this three-state area includes habitats ranging from coastal beaches,
dunes, broad estuarine tidal marshes and bald cypress swamps on the coastal plain to
upland forests and boreal bogs in the Appalachian Mountains. The area includes the
southernmost extent of the ranges of many northern species, the northernmost extent of
many southern species, and contains internationally significant migratory bird staging and
concentration areas. This area also includes the cities of Baltimore, Maryland;
Wilmington, Delaware; and Trenton, New Jersey; and is influenced by Washington, D.C.;
New York City; and Philadelphia, Pennsylvania. The region is heavily impacted by urban
5
development and suburban sprawl, and includes a large portion of the Delmarva
Peninsula which is significantly dominated by agricultural activities.
There is a diversity of topographic features from middle elevation mountains with a
maximum elevation of 1035 m (3395 ft) to sea-level barrier islands. There are 6 broad
physiographic provinces (Figure 1.2) of the 20 that occur in North America, each with a
mix of natural diversity and ecologically significant features. The mixed forests of the
Appalachian Plateau, Ridge and Valley, and Blue Ridge Plateau Provinces contain some
of the most diverse, ancient broadleaf forests on earth (Olson et al. 1998). The
Cranesville Sub-Arctic Swamp, a cool, “frost-pocket” bog, occurs along the western
boundary of Maryland’s panhandle, on the Allegheny Plateau.
New Jersey’s Piedmont Province is heavily developed, but still contains the remains of
several glacial lakes along with extensive freshwater wetlands. Approximately 25% of
the state is the protected Pinelands, a largely uninhabited area which includes Pine
Barrens (Walsh et al. 1999). Maryland’s Piedmont Province contains 769 ha (1900 ac) of
serpentine barrens in the Soldier’s Delight Natural Environment Area. Ninety-five
percent of Delaware lies in the Coastal Plain, with the Great Cypress Swamp occurring
along its southern boundary. Sixty-five percent of Delaware’s wetlands are inland
palustrine, freshwater and nontidal (Hess et al. 2000). All three states harbor numerous
examples of vernal pools throughout the Coastal Plain Province. These seasonally wet
depressions are environmentally sensitive habitats for a number of rare plants and
animals. One of the Coastal Plain’s great features is the Chesapeake Bay, the country’s
largest estuary which has a longer tidal shoreline than the State of California (Robbins
and Blom 1996). The Delaware Bay, an ancient, drowned river bed, separates Delaware
and New Jersey and facilitates traffic into Philadelphia, Pennsylvania which is one of the
busiest ports in the United States (Hess et al. 2000, Walsh et al. 1999).
6
Figure 1.1. Maryland-Delaware-New Jersey Gap Analysis Project study area
7
Figure 1.2. Physiographic Provinces of the Maryland-Delaware-New Jersey Gap Analysis Project study area
8
Chapter 2: Predicted Animal Species
Distributions and Species Richness
2.1 Introduction
All species range maps are predictions about the occurrence of those species within a
particular area (Csuti 1994). Traditionally, the predicted occurrences of most species
begin with samples from collections made at individual point locations. Most species
range maps are small-scale (e.g., 1:10,000,000) and derived primarily from point data to
construct field guides which are suitable, at best, for approximating distribution at the
regional level or counties for example. The purpose of the GAP vertebrate species maps
is to provide more precise information about the current predicted distribution of
individual native species according to actual habitat characteristics within their general
ranges and to allow calculation of predicted area of distributions and associations to
specific habitat characteristics.
GAP maps are produced at a nominal scale of 1:100,000 or better and are intended for
applications at the landscape or "gamma" scale (heterogeneous areas generally covering
1,000 to 1,000,000 hectares and made up of more than one kind of natural community).
Applications of these data to site- or stand-level analyses (site--a microhabitat, generally
10 to 100 square meters; stand--a single habitat type, generally 0.1 to 1,000 ha; Whittaker
1977, see also Stoms and Estes 1993) will likely reveal the limitations of this process to
incorporate differences in habitat quality (e.g., understory condition) or necessary
microhabitat features such as standing dead trees.
Gap analysis uses the predicted distributions of animal species to evaluate their
conservation status relative to existing land management (Scott et al. 1993). However,
the maps of species distributions may be used to answer a wide variety of management,
planning, and research questions relating to individual species or groups of species. In
addition to the maps, great utility may be found in the consolidated specimen collection
records and literature that are assembled into databases used to produce the maps.
Perhaps most importantly, as a first effort in developing such detailed distributions, they
should be viewed as testable hypotheses to be confirmed or refuted in the field. We
encourage biologists and naturalists to conduct such tests and report their findings in the
appropriate literature and to the Gap Analysis Program such that new data may improve
future iterations.
Previous to this effort there were no maps available, digital or otherwise, showing the
likely present-day distribution of species by habitat type across their ranges. Because of
this, ordinary species (i.e., those not threatened with extinction or not managed as game
animals) are generally not given sufficient consideration in land-use decisions in the
context of large geographic regions or in relation to their actual habitats. Their decline,
because of incremental habitat loss can, and does, result in one threatened or endangered
species "surprise" after another. Frequently, the records that do exist for an ordinary
9
species are truncated by state boundaries. Simply creating a consistent spatial framework
for storing, retrieving, manipulating, analyzing, and updating the totality of our
knowledge about the status of each animal species is one of the most necessary and basic
elements for preventing further erosion of biological resources.
There are three major data sets used in GAP to predict the distribution of vertebrate
species: 1) breeding ranges for all animal species; 2) a species-habitat association
database with tables that identify relationships between animal species and various habitat
variables; and 3) geographic information system (GIS) map overlays representing the
habitat variables for which species habitat relationships have been recorded in the
database tables.
2.2 Methods
The predicted animal species distribution mapping for Maryland, Delaware and New
Jersey began with the mapping of species’ ranges or distributional limits. Range maps for
most common species were based on confirmed or probable presence within the 650
square-kilometer hexagon units used by the Environmental Protection Agency’s
Environmental Monitoring and Assessment Program (EMAP). For most rare species, the
much smaller 7.5-minute quadrangle was used, primarily because this is one method
utilized by Natural Heritage Programs for tracking the distributions of rare species and,
therefore, data for these species were generally available at this scale. Although
information about the locations of some rare species is considered sensitive (e.g., for
collectible species such as the bog turtle), the use of smaller range units was preferred
because of the greater potential to overestimate distributions of rare species, many of
which are habitat specialists.
The habitat modeling component, which results in more precise mapping of predicted
animal species distributions within the range units, started with the compilation of habitat
relationships information from the literature. Using this information as a reference, a list
of commonly-described habitats (e.g., oak-hickory forest, salt marsh) was developed, and
other modeling variables (e.g., slope, aspect, elevation, distance from edge, proximity to
water) were identified. Raster-based modeling grids (i.e., map overlays) representing
these habitat variables were then developed and the habitat relationship information
gleaned from the literature was entered into an associated database of modeling tables.
2.2.1 Mapping Standards and Data Sources
All GIS modeling of species distributions was conducted in ArcView 3.2, controlled by
customized Avenue scripts, within a Windows 2000 operating system environment. Many
of the GIS map overlays used in the modeling were created in ARC/INFO version 7.1.2
on a Sun Workstation. All GIS overlays were developed as, or converted to, raster grids
with a 30-meter cell resolution, in the Universal Transverse Mercator projection (zone 18,
datum NAD83). The minimum mapping unit varied depending on the particular grid or
original data sources used to create grids. The GIS overlays (i.e., grids) used in the
10
modeling are listed in table 2.1, and more details about the development of individual
modeling grids are presented in the sections that follow the table.
Table 2.1: Grids Used in Habitat Modeling
MODELING GRID SOURCE DESCRIPTION
Range Extent or
Distributional Limits
by Hexagon
Biodiversity Research
Consortium, museum
records, other sources
Confirmed or Probable species
presence within 650 square-kilometer
hexagon range units
Range Extent or
Distributional Limits
by 7.5-minute
quadrangle
Natural Heritage
Programs, Breeding Bird
Atlas projects, other
sources
Confirmed or Probable species
presence within 7.5-minute quadrangle
range units
Habitat Types GAP Land Cover,
National Land Cover
Data, National Wetlands
Inventory, other sources
Source data sets were combined (see
section 2.2.3.1)
Wetland Buffer (100
m, 250 m, 500 m,
1000 m)
National Wetlands
Inventory; USGS
1:100,000 DLG
(streams)
NWI and DLG data were aggregated
into 14 wetland classes and buffered
(see section 2.2.3.2)
Forest Fragmentation
Metrics (Area, Patch
Isolation, Riparian
Forest Width)
National Land Cover
Data (NLCD)
ZONALTHICKNESS applied in
GRID to create Forest Area and
Riparian Forest Width grids;
FOCALMEAN applied to create patch
isolation grid, expressed as % forest
cover within 2 km (see section 2.2.3.3)
Open (Edge,
Grassland Area)
Habitat Type grid (see
above)
EUCDISTANCE applied in GRID to
calculate distance from forest/non-forest
edge; ZONALTHICKNESS
used to create Grassland Area grid (see
sections 2.2.3.4 and 2.2.3.5)
Land Form
(Elevation, Slope,
Aspect)
National Elevation Data
(30-m NED)
Elevation Z units are in meters; Slope
expressed as percent rise; developed in
Arc/Info GRID (see section 2.2.3.6)
Juxtaposition
(Roads, Forest)
USGS 1:100,000 DLGs
used for Road
Juxtaposition; Habitat
Type (see above) used to
develop Forest
Juxtaposition grid
Roads converted to raster grid and
EUCDISTANCE applied;
FOCALMEAN, with 250-m
neighborhood, applied to create Forest
Juxtaposition grid (see sections 2.2.3.7
and 2.2.3.8)
Special Habitat
Feature (island, cave,
outcrop, cliff,
dam/bridge)
Various (see section
2.2.3.9)
Each feature was buffered by 100
meters, 2 kilometers, 7 kilometers, and
15 kilometers
11
2.2.2 Mapping Range Extent
Existing range data sources for the MDN-GAP project included state Natural Heritage
Programs (NHP), museum records, study skin collections, and Breeding Bird Atlas
(BBA) projects. At the time that the range mapping was initiated, the Maryland BBA
project (Robbins and Blom 1996) was just being completed, and the Delaware and New
Jersey BBA projects were in the process of being completed (Hess et al. 2000, Walsh et
al. 1999). Data from these projects became available at different times and there were
associated delays in completing the range mapping. The data from these various sources
were used to develop the Biodiversity Research Consortium (BRC) data set, which is
based on the Environmental Protection Agency’s hexagons used in their Environmental
Monitoring and Assessment Program. Within the Maryland-Delaware-New Jersey
project area, these hexagons ranged in size from about 648 to 651 square kilometers per
hexagon. Because hexagons have a constant shape and size and are easily aggregated or
tessellated, they overcome many problems associated with delineating species ranges
using county boundaries (Boone 1996). The BRC effort was overseen by NatureServe,
with staffs from the NHPs, Maryland Department of Natural Resources (MDDNR), and
U.S. Fish & Wildlife Service (USFWS) involved in data gathering and development.
Although the Maryland and Delaware BRC projects were completed in draft form in
1997, there were erroneous records along the Virginia-Maryland border which were not
corrected until the BRC project was finalized in July of 2002. The New Jersey BRC
project was initiated much later, and was initially intended to cover only half of the state,
but with assistance from the USFWS, this project was extended to cover the entire state.
The New Jersey BRC data were made available in July of 2002, when the data sets for the
other states were finalized. The BRC dataset formed the basis for the range-mapping
component of the MDN-GAP.
The species records associated with each hexagon include a code indicating the level of
certainty of breeding occurrence for the species, as shown in Table 2.2. In general, only
those records with “probable” or “confident” levels of certainty were used. However,
there were cases where a hexagon with a “possible” level of certainty was surrounded by
hexagons with higher levels of certainty, and was therefore included in the modeling.
There were also cases where new information or personal knowledge provided
justification for inclusion of additional hexagons in a species range limits within the
project area. The BRC data were used for most common species, and for some rare
species, including three of the four modeled taxa (mammals, reptiles, amphibians) in New
Jersey, where availability of NHP data was limited. An example of a hexagon-based
range map is shown in Figure 2.1.
12
Table 2.2: Codes Indicating Level of Certainty of Species Breeding Occurrence in
Hexagon (Hernandez 2002)
LEVEL OF
CERTAINTY
EXPLANATION IN
NUMERICAL TERMS
BASIS FOR LEVEL OF
CERTAINTY OR EXAMPLES
Confident /
Certain
>95% certainty that the species
occurs in the hexagon -- species is
confidently assumed or known to
occur in the hexagon
recent, field-verified element
occurrence record in the heritage
database, museum record, or a
verified observation; the species’
habitat is believed still present in
the hexagon; and the species is not
a vagrant nor is it known to have
undergone any local decline that
would lead one to expect that it was
not still currently present
Predicted /
Probable
>= 80% certainty that the species
occurs in the hexagon -- species is
predicted to occur in the hexagon
based on the fact pattern (e.g.,
presence of suitable habitat or
conditions and historical record
and/or presence in adjacent
hexagon(s))
hexagon is well within the range of
the species and suitable habitat is
believed to be present but its
occurrence in the hexagon was not
known to be confirmed by the
developer of this data file
Possible 10%-80% estimated likelihood of
occurrence in the hexagon --
species possibly or potentially
occurs in the hexagon
hexagon occurs at the edge of the
species range, or the species is quite
rare and sporadically distributed
such that there is less than an 80%
probability that it is present in the
hexagon
For most rare, threatened, or endangered species, a separate range database was created,
with most records coming from the Natural Heritage Programs, and the smaller 7.5-
minute quadrangle unit was used. Natural Heritage Program data covering all of New
Jersey could not be obtained, so BRC data were used for all species in this state, with the
exception of rare, threatened, or endangered birds, for which BBA data were used to
populate quad-level records. Rules regarding levels of certainty of occurrence were
essentially the same in the Natural Heritage Program data and the BBA data. An example
of a quad-based range map is shown in Figure 2.2. Investigators for this project had
originally intended to run models at both the quad level and the hexagon level for all
species, in order to compare the results of the two approaches, but available resources for
this three-state project were inadequate to allow this extra level of effort. There was also
an interest in running bird models using the BBA blocks, each of which is one-sixth of a
7.5-minute quadrangle, but this extra initiative was also foregone due to inadequate
project resources.
13
Figure 2.1: Example of a Species’ Range by Hexagon
14
Figure 2.2: Example of a Species’ Range by 7.5-Minute Quadrangle
15
Due to the delays in completing the BRC projects for Maryland, Delaware and New
Jersey, some of the final revisions to the BRC data set were not incorporated into the
range data tables used in the modeling. However, all species ranges were reviewed
internally, and most, if not all, of the errors were discovered and corrected. In addition,
there are still some known problems with the final BRC data set that were addressed in
the modeling (e.g., range data for the red squirrel in the Coastal Plain of Maryland and
Delaware are considered erroneous). This internal review also led to the development of
“estimated” ranges for some, mostly common, species. Estimated ranges generally
included “possible” hexagon occurrences that were surrounded by hexagons with higher
levels of certainty of occurrence. However, in a few cases, hexagons were added based
on new information.
There were also examples of subspecies with differing habitat requirements which
necessitated separate models and then a merging of model results. For example, there are
two subspecies of the deer mouse, Peromyscus maniculatus, within the project area. One
subspecies, the woodland deer mouse, P. m. maniculatus, is generally restricted to
woodland habitats, while the other subspecies, the prairie deer mouse, P. m. bairdii, is
generally restricted to open, herbaceous habitats. Both subspecies occur within the
project area, but their ranges are not completely overlapping. Therefore, separate range
(hexagon) data were developed at the subspecies level, the two subspecies were modeled
separately, and the results were merged into a final species-level model. Similar issues
were addressed in much the same way for two subspecies of copperhead, Agkistrodon
contortrix, which has a northern subspecies that uses rocky habitats, and a southern
subspecies or intergrade that is found in swamps. There are also two subspecies of the
eastern earth snake, Virginia valeriae, one of which is a rare subspecies found only in the
mountains, and two subspecies of swamp sparrow, Melospiza georgiana, one of which is
found primarily in and around tidal marshes, while the other is found primarily around
inland, non-tidal marshes. Because the latter two subspecies have separate breeding
ranges within the project area, species-level modeling would have resulted in many errors
of commission. Although the National GAP standards and the BRC range data do not
support subspecies-level modeling, this extra level of effort was deemed necessary in a
few cases in order to achieve accurate model results.
2.2.3 Habitat Modeling Grids
2.2.3.1 Habitat Types
The primary species habitat modeling layer, one that was included in the modeling
equations of all species, was the Habitat Types grid, which was based on the GAP Land
Cover, National Land Cover Data (NLCD), and National Wetlands Inventory (NWI) data.
Authors who have identified and described wildlife habitat types in the eastern United
States include DeGraaf and Rudis (1986), Benyus (1989), DeGraaf et al. (1991), Hamel
(1992), and Robbins and Blom (1996). Many additional efforts have been made to
classify plant communities without regard for the vertebrates occupying the community.
These include Harshberger (1970), Brush (1975), Brush et al. (1980), the Society of
16
American Foresters (Eyre 1980), TNC in conjunction with state Natural Heritage
Programs (Sneddon et al. 1994; Sneddon and Berdine 1995; Clancy 1996; Clancy 1998;
Sneddon 1999), and the Federal Geographic Data Committee (FGDC) (1997). Additional
efforts have been focused on classifying natural communities, with consideration given to
both plant and animal communities (Kricher 1988, Breden 1989, Berdine 1998, Sneddon
1998). Cowardin et al. (1979) provide a classification of wetland and aquatic
communities based on plant species composition, hydrology, and other factors. Finally,
Anderson et al. (1976) have provided a classification of land use/cover types, including
urban and agricultural areas.
A key step in vertebrate distribution modeling is to provide a cross-walk from habitat
associations in the literature to land cover types generated in the land cover mapping
phase. We were constrained on several levels with regards to this objective. First, land
cover mapping was conducted concurrently with the vertebrate distribution modeling, and
land cover types were unavailable until late in the vertebrate modeling phase. Second,
very little of the available literature on species-habitat associations was specifically
focused on the mid-Atlantic region, and some sources that were focused on the mid-
Atlantic were not available until late in the vertebrate modeling phase (e.g., Walsh et al.
1999, Hess et al. 2000, Hulse et al. 2001,White and White 2002). Finally, many of the
sources did not consider the full range of potential habitat types available, but were
limited in their scope (forests and wetlands exclusively, for example).
As a consequence of these limitations, we chose to develop a standard list of wildlife
habitats (termed ‘Habitat Types’) for the project. They represent distinctions likely to
have unique assemblages of terrestrial and amphibious vertebrates, or a unique
combination of occupancy and utilization by terrestrial or amphibious vertebrates (i.e.
foraging, nesting, denning, overwintering, aestivation, etc.). Species’ responses to
environmental parameters in habitat selection vary from species to species, but key
parameters influencing distribution often include geographic context (latitude/longitude,
elevation, etc.), microclimate, plant community composition, vegetative structure, ground
conditions (leaf duff, soil type) and wetness (xeric, mesic, wetland hydrology).
Additional parameters might include wetland salinity, special habitat features (e.g., rock
outcroppings), and the degree of human disturbance. The habitat types were developed
with primary consideration given to these parameters and their effects on species
distributions. The steps taken in developing the final list of Habitat Types and their
descriptions were as follows:
1. A literature review was conducted of key sources representing authors who had
classified habitats or community types for the eastern U.S. based on either animal
communities (DeGraaf and Rudis 1986, Benyus 1989, DeGraaf et al. 1991, Hamel
1992) or plant and animal communities in combination (Kricher 1988). The
classifications they derived, including primary plant species composition, were
summarized in a document (Appendix B).
17
2. A spreadsheet of primary classifications from these sources was compiled. From this,
new categories were derived which captured similar classifications from multiple
authors. These ‘habitat types’ were named identically or with similar naming
conventions to source classification names. The spreadsheet is included in Appendix
C.
3. Aquatic habitat descriptions were developed based on modifications of Cowardin et al.
1979) and additional information from Tiner (1985), and urban and agricultural
habitats were modified from Anderson et al. (1976), based on known vertebrate use of
these areas.
4. Finally, the list was refined based on consultation with numerous other community
classification schemes, including Harshberger (1970), Brush (1975), Brush et al.
(1980), Eyre (1980), Breden (1989), Sneddon et al. (1994), Sneddon and Berdine
(1995), Clancy (1996), Robbins and Blom (1996), FGDC (1997), Berdine (1998),
Sneddon (1998), and Sneddon (1999). In addition, a partial crosswalk was developed
from the Habitat Types to TNC’s Alliances (Sneddon 1999), with reference to Gleason
(1963). While consulting these sources, numerous habitat types were added in cases
where identified plant communities had no previous representation in the Habitat
Types classification, but were very likely to support distinct animal communities. The
final list of 103 Habitat Types is included in Appendix D, and definitions are provided
in Appendix E. Crosswalks between many of the Habitat Types and Alliances are
available in Gorham and McCorkle (2006).
Once the list of habitat types was finalized, a table was built for use in cross-walking
GAP Land Cover classes or aggregations of classes into the Habitat Types. In reviewing
the draft GAP Land Cover as a part of this process, the decision was made to integrate
National Wetlands Inventory (NWI) data and National Land Cover Data (NLCD) into the
final habitat grid. This decision was based on several findings related to the GAP Land
Cover, among those being: 1) it included only two water classes, which would be
problematic for modeling certain species’ or animal groups’ distributions (e.g.,
amphibians), 2) there were forest classes that included both upland and wetland forests,
3) many wetland classes appeared to be under-mapped, compared with NWI, 4) many
areas known to be relatively pure hardwood forests were mapped as mixed forests, 5)
Atlantic white cedar swamps were found to be under-mapped in New Jersey, 6) bald
cypress swamps were mapped in New Jersey, where this swamp association does not
naturally occur, 7) water features larger than the stated minimum mapping unit were
missing from the Land Cover in some geographic areas, but were included in both the
NWI and NLCD, 8) steep slopes and cliffs along rivers were mapped as water in some
areas, 9) there was only one urban developed land use class, 10) certain special wetland
types that might potentially be derived from NWI, and that are very important to
particular animal communities, were not included (e.g., vernal pools), and 11) coastal
plain alliances or associations were mistakenly mapped in the mountains and montane
alliances or associations were mistakenly mapped on the coastal plain.
18
Because the draft land cover layer did not line up well with NWI, NLCD, or USGS
1:100,000 scale roads and hydrography, a third-order polynomial rubber sheet
transformation was applied using the WARP command in ARC/INFO GRID, using these
other data sets for control point links. NWI data were then aggregated into 32 wetland
classes corresponding with habitat types defined for this habitat layer. Extra steps were
needed for some wetland habitats, such as vernal pools which required selection of only
those wetlands that were isolated and had hydrology modifiers indicating at least seasonal
inundation, and, from this subset, further selection based on wetland size (area < 2 ha)
and shape (Patton Circularity Shape Index of <= 1.6). In addition, tidal wetlands with the
oligohaline modifier were lumped with freshwater tidal wetlands (also including riverine
tidal classes), and deciduous needle-leaved forest classes were assumed to be bald cypress
swamps on Delmarva and tamarack swamps in northern New Jersey. Finally, near-shore
estuarine and marine open water classes were defined as being within 300 m of shore,
with offshore classes being more than 300 m from shore. Once all wetland polygons
were reclassified to the habitat classes, the coverage was converted to a grid.
The NWI habitat grid had two-digit values and was multiplied by 1,000 to produce five-digit
values ending with three zeros. The NLCD grid also had two-digit values, and was
multiplied by 100,000 to produce seven-digit values ending in five zeros. The GAP Land
Cover grid had three-digit values, and was added to each of the above grids, producing a
grid having seven-digit values with the first two digits indicating the NLCD class, the
next two digits indicating the NWI class, and the final three digits indicating the GAP
Land Cover class. A cross-walk table was created and used for reclassifying the various
combinations of NWI, NLCD and GAP Land Cover. In general, the resulting habitat
class was determined by agreement between at least two of the input grids, but in cases
where there was no agreement, the default was generally the GAP Land Cover
classification. The primary objectives of this approach were to: 1) improve wetlands
mapping in the habitat grid, especially with regards to those wetlands that were excluded
from the GAP Land Cover as a result of the minimum mapping unit (e.g., vernal pools);
2) improve agreement between the resulting habitat grid and the wetland "buffer" (i.e.,
proximity) layers produced for the modeling (see section 2.2.3.2); 3) improve agreement
between the habitat grid and the forest fragmentation grids which were based on the
NLCD; 4) create distinct water habitat classes, since the GAP Land Cover had only two
water classes, and wildlife species respond differently to several different aquatic habitats
(e.g., pond, lake, lower perennial river, upper perennial river, tidal river, bay, ocean); 5)
make a distinction between upland and wetland classes sharing similar vegetation that
were lumped into one class in the GAP Land Cover; 6) better define wetland classes
based on the NWI hydrology modifiers (e.g., saturated versus inundated); and 7) create
additional distinctions in anthropogenic land uses. The cross-walk table referred to above
is too large to be included in the appendices of this report, but will be provided either as a
supplement to the final habitat modeling layer or may be obtained from the contact listed
in its metadata.
After the cross-walk-driven reclassification was completed, additional refinements were
required. For example, a physiographic province grid was used to create masks for
19
reclassifying GAP Land Cover classes which were inappropriately classified relative to
physiographic province (e.g., montane classes within the Coastal Plain). In addition,
aspect was used to reclassify various habitats. For example, on the coastal plain and
piedmont where the northern mixed forest habitat (containing hemlock) is rare except on
north-facing slopes (e.g., steep, north-facing slopes along the shores of the Chesapeake
Bay), any northern mixed forest habitat cell with an aspect between 45 and 315 degrees
(i.e., not north-facing) was reclassified to a different forest type – often mid-Atlantic oak-pine.
Aspect was also used to a limited extent to separate two other forest types: northern oak
and oak-hickory, with the former generally occurring on north- or east-facing slopes in
cooler, often more mesic conditions on deep soils, and the latter generally occurring on
south- or west-facing slopes in warmer, drier conditions on thinner soils. However, this
distinction was only deemed necessary for two GAP Land Cover classes that lumped both
forest types together: 1) “Red Oak-White Oak” which is described as being mesic to dry
and includes dry, acidic oak-hickory forests as well as northern aspect, mesic forests, and
2) “Mixed Oak-Sugar Maple” which is described as including stunted oak-hickory
woodlands on talus slopes with thin, dry, acidic soils, and oak-sugar maple forests on
deep, moist to well-drained loams and silt loams on north and east mid-slopes and coves.
Because these lumpings create problems from a wildlife habitat perspective, it seemed
appropriate to use aspect to separate them. Cells from these two Land Cover classes were
reclassified to the oak-hickory habitat type if they had an aspect between 135 and 260
degrees. If their aspect was between 280 and 360 degrees, or between 0 and 100 degrees,
they were reclassified to northern oak.
An elevation mask was also used to separate various habitats: Northern hardwood
generally occurs above 1000 meters in the mid-Atlantic; the mixed mesophytic forest
habitat generally occurs between 300 and 1000 meters; and the low-elevation mesic
hardwood habitat was defined as occurring below 300 m. A slope mask was used for the
high-elevation and mid-elevation woodland classes, which are defined as xeric
woodlands on steep, usually south-facing, slopes. Woodlands occurring on southern
aspects (135 to 260 degrees), on slopes greater than 100 percent, at elevations above 500
meters, were classified as high-elevation woodlands. Woodlands occurring within the
same slope and aspect ranges, but occurring at or below 500 meters, were classified as
mid-elevation woodlands.
Unclassified, isolated patches of water cells (i.e., that did not correspond with NWI and
were not contiguous with a classified aquatic habitat) were assigned unique values by
zone (i.e., contiguous patch of water cells) using REGIONGROUP, and were then
classified by size to either "lake" or "pond," based on the Cowardin (NWI) definitions for
these water classes. In general, an isolated patch of water greater than 8 hectares in size
was classified as a lake, and a patch less than 8 hectares was classified as a pond.
Unclassified water cells that were contiguous with classified aquatic habitats were dealt
with using a nearest-neighbor reclassification.
20
While oligohaline tidal marshes were lumped with freshwater tidal habitats, based on the
NWI oligohaline modifier, another approach was needed to separate salt marshes from
brackish marshes. Salinity maps for the Chesapeake and Delaware Bays were found in
Funderburk et al. (1991) and in Sutton et al. (1996), respectively. These maps were used
as a reference in creating salinity masks to separate salt and brackish marshes, with
brackish marshes ranging between 5 and 18 parts per thousand salinity, and salt marshes
ranging between 18 and 30 parts per thousand. Oligohaline marshes range between 0.5
and 5 ppt salinity.
A hemlock data set, created by the New Jersey Department of Environmental Protection
(NJDEP), was converted to a grid in the appropriate projection and used to select
corresponding forest. Where one or more of the three primary data sources (GAP Land
Cover, NLCD, NWI) indicated a conifer-dominated or mixed forest, the habitat was
classified as either Northern Conifer or Northern Mixed Hardwood - Conifer, both of
which are defined to include hemlock where the habitat occurs on a north-facing aspect or
in other cool, shaded situations (e.g., ravines). If the majority of the three primary data
sets indicated a hardwood-dominated forest, then the habitat was usually classified as
Low Elevation Mesic Hardwood, which is also defined to sometimes include hemlock, as
long as the elevation criterion was met.
Feedback from a New Jersey GAP research associate indicated that Atlantic white cedar
swamps were under-mapped in the GAP Land Cover. An Atlantic white cedar swamp
data set, also created by the NJDEP, was converted to a grid in the appropriate projection
and used to select corresponding forest. Where one or more of the three primary data
sources (GAP Land Cover, NLCD, NWI) indicated a conifer-dominated or mixed
forested wetland, the habitat was classified as Atlantic White-Cedar Swamp.
There was also a slope-related issue which was discovered in the western Maryland GAP
Land Cover. Cliff shadows along the Potomac River were classified as water, and NWI
was used to more accurately define the river’s extent in this area. The remaining cells
were reclassified to the “cliff” habitat type, except where the NLCD provided vegetated
classes which were classified to various steep-slope vegetated habitat types.
Prior to finalizing the Habitat Types grid, unresolved cells were reselected and any
contiguous clusters of 5 or more cells (0.45 ha) were identified using REGIONGROUP.
These clusters were reevaluated and classified to the most appropriate habitat type. Once
these clusters were classified, a nearest neighbor classification was applied to the
remaining, unclassified cells. A map of the Habitat Types in New Jersey is shown in
Figure 2.3.
21
Figure 2.3: Habitat Types in New Jersey
22
Of the 103 habitat types which were defined for this project, several were not mapped for
various reasons. For example, sparsely vegetated habitats such as "outcrop" and "gravel
barren" were generally not mapped because these classes were not captured in the GAP
Land Cover. Cliff data became available after this habitat grid was finalized. There are
many cells which should be mapped as the cliff habitat type, but are mapped as other
types. Although the "seep" habitat type is thought to be important for several amphibian
species, this was not mapped because it generally occurs as a very small feature on the
landscape and it could not be derived from the GAP Land Cover or other ancillary data.
Some habitat types were not defined but, in retrospect, should have been defined and
mapped (e.g., impoundments, aquatic beds). With regards to minimum mapping unit,
this data set is relatively good in terms of completeness. NWI data were used to capture
vernal pools and farm ponds as small as 0.09 hectare (0.22 acre; one 30-m cell), which
were otherwise smaller than the minimum mapping unit of the GAP Land Cover. A
possible drawback to this is the earlier vintage of the NWI (generally 1980s), which may
have led to some errors of commission where such features have been lost through
development or conversion to agriculture, but such errors were generally avoided where
both the GAP Land Cover and the NLCD indicated an anthropogenic land use class.
A very important habitat which could not be included in the habitat layer was the
"stream" habitat type, since most streams are much narrower than a 30-m cell. If NWI
and USGS mapped a water feature as a polygon, then it was included in the habitat layer,
but if the water feature was captured only as a linear (non-polygonal) feature in both of
these data sets, then it could not be included in the habitat layer. This necessary omission
was compensated for by a separate wetland/water feature buffer (proximity) modeling
layer which is described below. Finally, the NLCD developed by EPA was used to add
small woody habitats (i.e., smaller than the 2-ha minimum mapping unit of the GAP Land
Cover) to the habitat layer, since these habitats are important to edge species. These cells
were generally classified as Mid-Successional Old Field since they were mostly disturbed,
edge habitats.
2.2.3.2 Wetland Buffers
To some degree, many animal species are associated with wetlands. Some species are
almost always found near wetlands, and studies of certain species groups indicate
predictable numerical relationships. For example, adult salamanders (n = 265) of six
species (Ambystoma jeffersonianum, A. maculatum, A. opacum, A. talpoideum, A.
texanum, A. tigrinum) were found an average of 125.3 m from the edge of aquatic
habitats during the non-breeding portions of their life-cycles, and a wetland buffer zone of
164.3 m (534 ft) could be expected to encompass the majority of the population of these
salamanders during their entire life cycle (Semlitsch 1998). The spotted turtle (Clemmys
guttata) is generally found within 500 m of a wetland (Whitlock 1994). Gardner (1982)
stated that the Virginia opossum (Didelphis virginiana) requires considerable amounts of
water to avoid dessication, and accessibility of surface water may be critical to suitable
opossum habitat. Sandridge (1953) found that the greatest distance between any opossum
den and a source of drinking water was approximately 366 m (1,200 ft) [In Gardner
1982]. In a study of the habitat requirements of the osprey (Pandion haliaetus), Ewins
23
(1997) found that 93% of 179 tree nests were within 500 m of water, and the median
distance to water for tree nests was 10 m (vs. 4 m for nests on artificial platforms) [In
Poole et al. 2002].
In some cases, numerical data are not provided, but authors state that a species is
generally found “close to streams,” “along stream margins,” “along swamp margins,” or
“in floodplains.” In these cases, knowledge of the species’ home range size was used in
assigning the species to one of four wetland buffer distances. The four “buffer” distances
chosen for inclusion in modeling the habitat requirements of species that most commonly
occur near wetlands were 100, 250, 500, and 1000 m. In addition, fourteen general
wetland types were identified as being important to one or more species: 1) stream, 2)
river (both tidal fresh and non-tidal), 3) lake, 4) pond, 5) swamp (forested), 6) shrub
swamp, 7) saturated/temporary, 8) vernal pool, 9) fresh marsh (non-tidal), 10) fresh tidal
marsh, 11) salt/brackish marsh complex, 12) estuarine river/stream/pond, 13) salt bay,
and 14) ocean. A table of species-wetland buffer relationships was created for each of the
four taxa (birds, mammals, reptiles, amphibians), and four “hypergrids” were created, one
for each buffer distance, by combining the buffers of the 14 wetland types according to
the following methods:
NWI served as the primary data source for developing this modeling layer. Wetlands
were aggregated into most of the types listed above based on NWI codes (see Cowardin et
al. 1979) which indicate wetland SYSTEM (e.g., estuarine), SUBSYSTEM (e.g.,
intertidal), CLASS (e.g., emergent), and, in some cases, SPECIAL MODIFIERS (e.g.,
oligohaline). In addition, the Patton Circularity Shape Index was calculated for certain
palustrine wetlands in order to develop a subset of wetlands meeting one of the identified
criteria for vernal pools. Other criteria for vernal pools included size (area < 2 ha), and
hydrology (NWI hydrologic modifiers indicating at least seasonal inundation). All of the
wetland buffer types listed above were derived from NWI, with the exception of the
“stream” wetland type, which was created from USGS DLGs (see below). The resulting
wetland coverage was converted to 13 separate grids, one for each wetland type. The
EUCDISTANCE command was then applied to each GRID, to buffer the wetlands to
each of the four buffer distances (100 m, 250 m, 500 m, 1 km), creating four separate
grids for each of the 13 wetland types. This approach is cleaner than buffering polygons
in a vector format.
The USGS 1:100,000 Hydrography data were used to develop the stream component of
the wetland buffer grids. Using NWI, a "salt mask" was created, which was essentially a
polygon that included all estuarine tidal wetland areas, but excluded those with the
oligohaline modifier. This polygon was intersected with the preliminary stream coverage,
and all stream segments occurring within that area were deleted, leaving just those stream
segments outside of the saltwater tidal areas. The final stream coverage was buffered to
the four buffer distances, and these coverages were converted to grids. The stream
segments that fell within the salt mask were also buffered and converted to grids, as were
NWI line features falling within this zone, and the resulting grids were merged with the
Estuarine River/Stream/Pond wetland buffer grids created in the previous step.
24
The final Wetland Buffer modeling layers were created by combining the individual
component grids (stream, river, lake, pond, swamp, shrub swamp, saturated wetland,
vernal pool, fresh marsh, fresh tidal marsh, salt/brackish marsh, estuarine
river/stream/pond, salt bay, and ocean), each buffered to four distances (100 m, 250 m,
500 m, 1 km) for a total of 56 separate buffer grids, into 4 binary-coded "hypergrids," one
for each buffer distance, such that the placement of the character in the binary code
denotes the wetland type. An AML, written by Jason Karl (Idaho Cooperative Fish and
Wildlife Research Unit) for use in combining final species models into multiple-species
hypergrids, was used to combine the different wetland buffers into the hypergrids.
It should be noted that, for all modeling variables, a control table determined whether or
not a particular modeling variable was “required.” If a variable was required (e.g.,
species is restricted to habitats that are within 100 meters of a particular wetland type),
then the final mapped species distribution was “clipped” by that variable. Conversely, if
the control table indicated that a particular variable was not required by the species, then
portions of the species’ distribution influenced by that variable might receive a higher
overall suitability ranking in the final results, but the species’ distribution would not be
excluded from areas outside of the influence of that variable.
2.2.3.3 Forest Fragmentation Variables
The conservation of birds requires an understanding of their nesting requirements,
including area as well as structural characteristics of the habitat (Robbins et al. 1989).
Several studies have shown that many bird species seem to depend on extensive forested
areas to support viable breeding populations. (Robbins et al. 1989, Keller et al. 1993,
Kilgo et al. 1998, Whitcomb et al. 1981, Lynch and Whigham 1982, Anderson and
Robbins 1981, Robbins 1979), and forest area requirements have been summarized by
various authors (Hamel 1992, bushman and Therres 1988, Rosenberg et al. 1999).
Species that appear to be sensitive to forest fragmentation are sometimes referred to as
forest interior-dwelling (FID) species or forest area-dependent (FAD) species. There are
some species that are sensitive to forest patch isolation, requiring a large amount of
overall forest cover, but which do not necessarily require forest interior. Therefore, the
latter of the two terms is more applicable to this aspect of the modeling.
FAD species were defined as species showing a significant (p < .05) negative response to
forest fragmentation in one of any number of published studies conducted in the eastern
United States. The typical research approach and analysis in studies of this nature
involves breeding season point counts or transects, detailed measurement of vegetation
and other environmental variables, including fragmentation metrics, at point count
locations, and analysis including stepwise multiple regression to identify which
environmental variables are significant predictors of nesting occurrence.
Modeling FAD species distributions required the development of three forest
fragmentation data layers, based on metrics identified as significant in published studies.
These were forest patch size measured by zonal thickness, riparian forest width, and the
25
percent of forest within 2 km as a measure of forest patch isolation. These metrics are
illustrated in Figure 2.4.
Figure 2.4: Forest Fragmentation Metrics used in Habitat Modeling.
Suitability of values in the fragmentation layers for each species was determined on a
species by species basis from probability curves output from logistic regression analysis
(see Figure 2.5). Data from two primary studies, Robbins et al. (1989) and Keller et al.
(1993), were used for this process. The latter study was used for riparian dependent
species, and the former for other species. Probability curves are species specific, with the
x axis on these curves representing the fragmentation metric, and the y axis representing
the probability of occurrence for that species. Fragmentation metric values corresponding
with 80% of the maximum occurrence of a species were considered optimal, values
corresponding with 50% of the maximum were considered suitable, and values
corresponding with 20% were considered marginal. Values less that 20% of the
maximum were not considered habitat. Table 2.3 provides a summary of the
fragmentation metrics and the suitability thresholds used on a species by species basis.
26
Figure 2.5: Example of Probability Curve (Robbins et al. 1989).
ZONALTHICKNESS is an ARC/INFO GRID function which measures the radius of the
largest circle that will fit within a zone, in this case a forest patch. This was used as a
surrogate for forest patch size because it provided an automated way to reduce the forest
interior value of irregularly shaped patches or long linear forests; these forest patches
were manually eliminated in the published studies we evaluated. A calibration of zonal
thickness to the forest patch size as determined in the field studies was conducted from
records of the original point locations (Figure 2.6).
Table 2.3: Modeling Parameters and Suitability Thresholds for Area Sensitive
Species
Significant
Modeling
parameters
(P<.05) variable minimum mid range high range
SPECIES (any study) used (>marginal) (>suitable) (>optimal)
Red-shouldered hawk yes IS2 37.2 71.1 90.1
Barred owl RIP 188.3 580.8 1159.9
Pileated woodpecker yes LAR 11.6 164.9 974.5
27
Significant
Modeling
parameters
(P<.05) variable minimum mid range high range
SPECIES (any study) used (>marginal) (>suitable) (>optimal)
Hairy woodpecker yes LAR 1.4 6.5 367.1
Acadian flycatcher yes LAR 0.2 14.7 389.8
Yellow-throated vireo yes IS2 36.6 69.9 89.5
Red-eyed vireo yes LAR 0.3 2.3 16.2
White-breasted nuthatch yes LAR 0.5 1.5 193.9
Brown creeper yes IS2 58.4 81.5 93.9
Blue-gray gnatcatcher yes LAR 0.8 13.7 452.7
Veery yes LAR 4.1 49.6 712.3
Wood thrush yes LAR 0.2 0.2 26
Northern parula yes LAR 65 528.3 1674.6
Black-throated blue warbler yes LAR 523.3 1079.3 1630.7
Cerulean warbler yes LAR 115.8 713.9 1872.9
Black-and-white warbler yes LAR 12.2 224.8 1219.4
American redstart yes IS2 15.8 61.9 87.2
Prothonotary warbler yes RIP 121.8 261.7 562.6
Worm-eating warbler yes LAR 5.8 153.2 1055.4
Swainson's warbler yes RIP
Ovenbird yes LAR 0.8 9.1 232.9
Northern waterthrush yes LAR 16.7 190 855.8
Louisiana waterthrush yes RIP 121.3 262 580.8
Kentucky warbler yes RIP 5.3 47.3 716.5
Hooded warbler yes IS2 14.6 58.9 85.4
Canada warbler yes LAR 56.2 369.8 1116.2
Summer tanager yes LAR 0.8 47.4 736.1
Scarlet tanager yes LAR 0.9 12 128.8
Rose-breasted grosbeak yes LAR 1.1 1.1 88
LAR - area of forest stand (ha) as modeled by Robbins et al. (1989)
IS2 - forest isolation measured as % forest within 2 km radius as modeled by Robbins et
al. (1989)
RIP - riparian forest width as modeled by Keller et al. (1993)
minimum: area/percent/width where modeled frequency of detection = 20% of maximum
(marginal 20-49%)
mid range: area/percent/width where modeled frequency of detection = 50% of max.
(suitable 50-79%)
high range: area/percent/width where modeled frequency of detection = 80% of max.
(optimal 80-100%)
28
Figure 2.6: Correlation of Zonal Thickness and Natural Log of Forest Area as
determined in Robbins et al. (1989).
The first step in developing the forest area modeling grid was to select forest classes and
other woody classes from the NLCD, and apply various processes and filters to the data in
order to: 1) eliminate small forest openings (< 1ha) not considered substantial enough to
affect FAD species occurrence, and 2) separate forest patches tenuously connected so
they would be considered separately in zonal thickness analysis. USGS class 1 and 2
(major) roads data were also used to separate tenuously connected forests. The selected
line coverage for major roads was converted to a grid, merged with the forest grid, and
then set to NODATA to create this separation. Secondary and other minor roads were
assumed to be insignificant in terms of breaking the continuity of a forest patch.
Although the distinction between major and minor roads is somewhat arbitrary and
subjective, it was driven by a preliminary evaluation of the forest patch grid in which
forest patches that appeared to be separate and distinct, and were bisected by major
highways, were nevertheless tenuously connected in the NLCD. By comparing bird
populations in forests on both sides of power-line and road corridors of different widths,
29
Robbins et al. (1989) determined that gaps of 100 m or more produced isolation
characteristics in the small fragments created.
After applying the major roads grid to achieve some separation of forest patches, the
SHRINK command was used in GRID to create further separation between patches.
Next, two filters (majority filter and focal majority) were applied to eliminate small (e.g.,
single-cell) openings in the canopy, essentially smoothing the forest patch grid in order to
obtain more accurate zonal thickness (i.e., forest patch depth) measurements. These
processes are described in greater detail in the metadata that accompanies this modeling
grid. Once the filters were applied, the EXPAND command was applied to expand the
forest patches back to their original sizes.
After the forest data were smoothed and tenuously-connected patches were separated,
REGIONGROUP was used to assign each spatially distinct forest patch a unique value.
This allows the final processing step, measurement of zonal thickness, to evaluate each
distinct patch separately. Prior to this final step, a mask was applied to eliminate distinct
patches having a count of less than or equal to 10 (i.e., less than 1 ha), including forest
canopy openings below this threshold. Such openings would generally be less than 100
m wide, regardless of shape. The ZONALTHICKNESS measurement was then used to
measure the maximum depth into a forest patch. A map depicting forest area as measured
by ZONALTHICKNESS is shown in Figure 2.7.
The width of riparian forests was also determined from zonal thickness analysis, which
was applied to all forests adjacent to wetland or water features. In this case, the radius of
the largest circle becomes a direct measure of one-half the width of the riparian forest.
For the forest patch isolation modeling layer, the chosen metric was based on the
approach used by Robbins et al. (1989), where patch isolation is related to percentage of
forest cover within 2 kilometers of the site being evaluated. After reclassifying NLCD to
forest (value = 100) and non-forest (value = 0), a FOCALMEAN process was run in
GRID in order to develop this modeling layer. This process measured the percentage of
forest cover within a 2-kilometer radius of each grid cell. A map depicting forest patch
isolation in Delaware is shown in Figure 2.8.
30
Figure 2.7: Map Depicting Forest Area Metric
31
Figure 2.8: Forest Patch Isolation in Delaware
32
2.2.3.4 Open - Grassland Area
Just as many forest-dependent birds are area-sensitive, many grassland birds also require
large, contiguous habitat patches to maintain viable breeding populations. Habitat area
requirements for grassland birds were taken from several studies (Jones and Vickery
unpubl., Swanson 1996, Samson 1980, Smith 1992, Smith 1991, Herkert 1994b, Herkert
1991) and minimum suitability thresholds were defined for each species. The process by
which the grassland area modeling grid was created was essentially the same as that used
to create the forest area grid. The herbaceous habitats evaluated included herbaceous old
field, upland riparian herbaceous, maritime grassland, wet meadow, fresh marsh,
herbaceous vernal pool, fresh tidal marsh, brackish marsh, low salt marsh, high salt
marsh, maritime marsh, forb-like crop, grass-like crop, pasture, clear-cut, and agricultural
barren / fallow. Note that, although many of these habitats are not generally used by
grassland species, they would not constitute “breaks” in grassland area where they are
contiguous with appropriate grassland habitat, and unsuitable habitats would be
eliminated as a result of the “habitat type” selection part of the modeling. The northern
harrier is known to be area-sensitive and prefers high marsh habitats.
2.2.3.5 Open - Edge Habitat
While some species require large, contiguous patches of habitat, far away from edges,
other species prefer edges. For these species, an Edge habitat grid was created. This
involved first reclassifying all woody habitats into one class and all non-woody habitats
into another class. A EUCDISTANCE process was then applied to each, separate class,
with a specified maximum distance of 300 meters. This upper threshold was based on a
study that found that nest parasitism by brown-headed cowbirds decreased with distance
away from forest edge, but extended >= 300 meters into the forest (Brittingham and
Temple 1983). Based on this and other information, it was decided that a distance of 300
m, extending in both directions away from an edge, should encompass most of the
activities and habitat needs of "edge" species. Once Euclidean distance was applied to
both grids (woody and non-woody habitats), the two results were merged.
2.2.3.6 Land Form (Elevation, Slope, and Aspect)
Elevation, Slope and Aspect are also important variables for determining the distributions
and preferred habitats of some species. These modeling grids were derived from the
National Elevation Data (NED) set. Elevation is expressed in meters. Because the NED
has a 30-m cell resolution, elevations were averaged over a 900 square-meter area for
each cell. Therefore, slope is based on the relationships among cells with averaged
elevation values, and this data set is only accurate for coarse-scale analyses (e.g.,
1:100,000-scale or greater). The DEMGRID command was used in ARC/INFO GRID, to
create the elevation grid. The SLOPE command was used in GRID, with the
PERCENTRISE option, to create the slope grid, and the ASPECT command was used to
create the aspect grid, which has values ranging from 0 to 359 degrees.
2.2.3.7 Road Juxtaposition
For a small number of species, studies have indicated a negative response to roads and a
positive correlation with distance from roads (Clark et al. 1993, Gibbs 1998). A road
33
juxtaposition grid was developed for use in modeling these species’ distributions. USGS
1:100,000-scale roads were appended into a seamless coverage for the project area, all
road classes except for class 5 (trails) were selected and converted to a 30-m grid, and the
EUCDISTANCE command was used in GRID to create a grid depicting road proximity.
The value for each cell in this grid represents the distance of the cell from the nearest
hard-surfaced road.
2.2.3.8 Forest Juxtaposition
There are many animal species that can be found in open, non-forested habitats during
some part of their life cycle or while meeting some life history requirement, but are
generally found in close proximity to forest and depend on forest habitats for meeting
some of their needs. For these species, a forest juxtaposition modeling grid was created.
This grid was initially created with mole salamanders in mind. These salamanders,
belonging to the genus Ambystoma, require upland forest habitat during the non-breeding
portions of their life cycles, when they spend most of their time in underground burrows,
under logs, and in moist leaf duff. They generally require relatively closed canopy
conditions, high ground-level moisture, and the presence of leaf duff and coarse woody
debris in various stages of decomposition.
Because different forest associations exhibit these characteristics to different degrees
(e.g., northern oak vs. coastal plain pine), the first step in developing this modeling layer
involved creating a system for ranking different forest types for their ability to satisfy the
requirements of these salamanders. A table was developed for ranking all woody habitats
based on four characteristics: 1) canopy closure, 2) coarse woody debris, 3) leaf duff, 4)
moisture (see Table 2.4). These rankings were subjective, but considered necessary since
some woody habitats meet the non-breeding habitat requirements of these species better
than others. Woody habitats received scores between 0 and 100, with 100 representing
optimal forest conditions. Non-woody habitats were assigned a value of 0. The Habitat
grid was then reclassified, according to this ranking system. Because a broad range of
conditions may be aggregated into a particular habitat type, none of the woody habitats
received an optimal ranking, although this aspect of the modeling may need revisiting.
Table 2.4: System for Ranking Salamander Non-Breeding Habitat
Note that all herbaceous and anthropogenic habitats (with the exception of PLANTATION and
CLEARCUT) were assumed to have no value as non-breeding habitat for the subset of species for which
this habitat modeling variable was developed (i.e., mole salamanders, other forest-dependent amphibians).
Although this is a very subjective ranking process, based on habitat descriptions, it is still preferable to
treating all woody habitats as equally good, in terms of meeting the non-breeding habitat needs of these
species. Some summer draw-down and/or microtopographic diversity in wetlands is assumed, and ranking
considers a range of conditions lumped into each habitat type.
HT_CODE HABITAT TYPE CWD DUFF MOIST CANOPY AVG
UF.BOCO BOREAL CONIFER 50 25 50 50 44
UF.BOHA BOREAL HARDWOOD 75 75 50 50 63
UF.BOMI BOREAL MIXED HARDWOOD-CONIFER 75 50 50 75 63
UF.NOCO NORTHERN CONIFER 50 25 50 75 50
UF.NOOK NORTHERN OAK 100 75 50 100 81
UF.NOOC NORTHERN OAK-CONIFER 100 50 50 100 75
34
HT_CODE HABITAT TYPE CWD DUFF MOIST CANOPY AVG
UF.NOHA NORTHERN HARDWOOD 100 75 50 100 81
UF.NOMX NORTHERN MIXED HARDWOOD-CONIFER 75 50 75 100 75
UF.MIME MIXED MESOPHYTIC 100 75 75 100 88
UF.APCO APPALACHIAN COVE HARDWOOD 100 75 75 100 88
UF.PIBA PINE BARREN 50 25 25 50 38
UF.OKHK OAK-HICKORY 100 75 50 100 81
UF.MAOP MID-ATLANTIC OAK-PINE 75 50 50 75 63
UF.LEMH LOW ELEVATION MESIC HARDWOOD 100 100 75 100 94
UF.CPPI COASTAL PLAIN PINE 50 25 50 75 50
UF.CPPO COASTAL PLAIN PINE-OAK 75 50 75 100 75
UF.HEWL HIGH-ELEVATION WOODLAND 50 25 0 50 31
UF.MEWL MID-TO LOW-ELEVATION WOODLAND 50 50 25 50 44
UF.MTFW MARITIME FOREST/WOODLAND 25 25 25 50 31
WF.BOFO BOG FOREST 25 25 75 50 44
WF.BOSP BOREAL SWAMP 50 25 75 50 50
WF.NCSP NORTHERN CONIFEROUS SWAMP 50 25 75 50 50
WF.NHSP NORTHERN HARDWOOD SWAMP 75 25 75 75 63
WF.AWCS ATLANTIC WHITE-CEDAR SWAMP 50 25 75 50 50
WF.CYSP BALDCYPRESS SWAMP 75 25 50 50 50
WF.BHSP BOTTOMLAND HARDWOOD SWAMP 75 25 75 75 63
WF.DSPH DEEP SWAMP HARDWOOD 75 25 50 50 50
WF.CPPF COASTAL PLAIN PINE FLATWOOD 50 25 75 50 50
WF.OKSP MIXED OAK SWAMP 75 25 75 75 63
WF.PHSP COASTAL PLAIN PINE-HARDWOOD SWAMP 50 25 75 75 56
WF.NORI NORTHERN RIPARIAN 75 25 75 75 63
US.ABHT ALPINE/BOREAL HEATH 0 25 25 25 19
US.KRUM KRUMMHOLZ 25 25 25 25 25
US.MHTB MONTANE HEATH THICKET/BALD 0 25 25 25 19
US.SSOF SHRUB/SAPLING OLD FIELD 25 25 25 25 25
US.MSOF MID-SUCCESSIONAL OLD FIELD 50 50 25 50 44
US.PBSC PINE BARREN SCRUB 25 25 0 25 19
US.DMTS DUNE / MARITIME THICKET / SHRUB 0 0 0 25 6
WS.NBBO NORTHERN/BOREAL BOG 25 25 50 25 31
WS.NBFE NORTHERN/BOREAL FEN 25 25 50 25 31
WS.SMSS SALT MARSH SCRUB 25 0 25 25 19
WS.MWTS MARITIME WET THICKET/SHRUB 25 25 50 25 31
WS.WVPO WOODY VERNAL POOL 50 50 75 50 56
WS.SSSP SATURATED SHRUB SWAMP 25 25 75 25 38
WS.FSSP FLOODED SHRUB SWAMP 50 25 50 25 38
WS.RITS RIPARIAN THICKET/SHRUB 25 25 75 25 38
AN.APLA AGRICULTURAL PLANTATION 25 0 25 50 25
AN.ARCL AGR. REGENERATING CLEARCUT 50 50 25 25 38
CWD = RELATIVE AMOUNT OF COARSE WOODY DEBRIS IN HABITAT
DUFF = RELATIVE AMOUNT OF DECIDUOUS LEAF DUFF ACCUMULATION
MOIST = RELATIVE MOISTURE AT GROUND LEVEL (MOIST, BUT NOT WET, OPTIMAL)
CANOPY = RELATIVE AMOUNT OF CANOPY / SHADE
AVG = AVERAGE RATING (CWD + DUFF + MOIST + CANOPY) / 4
35
A FOCALMEAN process was then applied to the reclassified Habitat grid. This process
assigned to each cell a value representing the average value for all cells within a 240-
meter (8-cell), circular neighborhood. This radius was a compromise between the
terrestrial life zone (zone surrounding amphibian breeding habitat such as a vernal pool)
requirement recommended by Semlitsch (1998) and the often-cited, more generous
upland forest buffer requirement of 250 meters. Note that the Semlitsch recommendation
of a 164-meter buffer zone is expected to encompass 95% of vernal pool-breeding
amphibians, but was thought to be an underestimate for some species (e.g., eastern newt,
Notophthalmus viridescens). In the modeling, this forest juxtaposition grid causes a
vernal pool in the middle of a farm field to get a lower suitability ranking than that of a
vernal pool in the middle of a hardwood forest.
Although this grid was developed primarily for use in modeling the habitats and
distributions of vernal pool-breeding salamanders, it was included in the models of
several other species that use non-forested habitats but are generally found in close
proximity to forests. For these species, the bias toward certain forest types was taken into
consideration, and this modeling variable was appropriately weighted in the modeling
equation such that this bias would not have an inappropriate influence on the final results.
2.2.3.9 Special Habitat Features
In addition to demonstrating an affinity for certain plant communities, land form
characteristics that influence these communities, and juxtaposition of habitats, there are
also special habitat features that many animal species use or require. Some of these
features cannot be included in landscape-scale mapping (e.g., nest cavities or boxes),
while others can be mapped at such scales if data are available. Of the many special
habitat features identified, only five were included in the final modeling: 1) island, 2)
cave, 3) outcrop, 4) cliff, and 5) dam/bridge. There were other special habitat features
that were considered important and mappable, including shale barrens and vertical stream
banks (for bank swallow colonies), but data could not be obtained in time for inclusion in
the modeling. Four buffer distances, 100 m, 2 km, 7 km, and 15 km, were chosen to
cover the range of distances found in the literature for species that use these features.
2.2.3.9.1 Island
The Island special habitat feature is important for colonial-nesting herons, egrets, gulls
and terns, which often nest most successfully on islands where human disturbance and
predation are minimized. An Island data set was created by the Maryland Department of
Natural Resources (MDDNR), but it covered only the Maryland portion of the three-state
project area. This data set was created from National Wetlands Inventory data and
personal knowledge. Vegetated wetland and upland polygons surrounded by water were
selected to create this data set. Additional islands were similarly selected for Delaware
and New Jersey.
2.2.3.9.2 Cave
A Caves (and mines) point coverage was provided by MDDNR, Wildlife and Heritage
Division, along with criteria for evaluating the suitability of each cave for meeting the
36
habitat requirements of bat species that depend on these special features. MDDNR also
obtained New Jersey cave data, and added these points to the data set. Not all caves in
the point coverage were considered suitable habitat for species that use caves. The
database associated with the point coverage included comment fields and other fields that
evaluated caves in terms of elevation, mineral type (e.g., limestone, marble, sandstone,
dolomite, shale, etc.), access (i.e., does the cave have an opening to allow wildlife
access), length (e.g., cave length is positively correlated with bat use), air flow (indicates
two or more entrances, complexity, chimney effects, and generally required for bat use),
and known bat use. Cave suitability variables were based on Raesly and Gates (1987)
and Navo (1994).
The variables and the scores given for each variable are shown in table 2.5. The scores
were tallied for each cave to select a final subset of caves to be buffered and used in the
habitat modeling for bats and other cave-dependent species. The highest possible score
was 10, and the score was divided by 10 to obtain an index.
Table 2.5: Variables used in evaluating suitability of caves for bat use
VARIABLE SCORE
Passage Length < 100’ 1
100-700’ 2
700-1100’ 3
1100-2400’ 4
> 2400’ 5
Mineral Type Soft Rock 1
Hard Rock 2
Air Flow Yes 1
No 0
Known Bat Use Yes 2
No 0
The final subset of caves included in the Special Habitat Features layer included only
those caves with a suitability index of >= 0.5, with one exception -- a cave having a score
of 0.4 that has water, supports a salamander population, and is rich in invertebrate fauna
(note that the cave buffer component of the Special Habitat Features layer was also used
in modeling the habitats of a few salamander species that are associated with caves).
2.2.3.9.3 Outcrop
Outcrop data were not available, so all caves and mines, including those that did not meet
the cave criteria, are included in this coverage, even though some may not have
corresponding outcrops. Most of the species associated with outcrops are responding
more to the presence of subterranean habitats associated with these outcrops than they are
to the surface of the outcrop. The assumption is that where there are caves or mines,
37
there are also likely to be rock outcrop formations. However, it is recognized that the
caves data set is a poor substitute for an accurate accounting of outcrops and that this
surrogate includes only a subset of outcrops found in the project area.
2.2.3.9.4 Cliff
Initially, no cliff data were available, so an analysis was undertaken to compare known
cliff locations with slope data. It was determined that all known cliffs (e.g., those named
on topographic maps) were associated with slopes >= 110% in the NED-derived slope
data. Grid cells associated with slopes < 110% were reclassified to nodata, and the
remaining grid cells were reclassified to zero, to create a preliminary cliff layer, which
became the final cliff layer for New Jersey. A comparison of this final data set with
known cliff locations along the Hudson River and upper Delaware River indicates a
reasonably accurate result. Cliff data for western Maryland became available later in the
project, through the Ecological Land Unit (ELU) data set created by The Nature
Conservancy. ELUs are unique combinations of three primary factors (elevation,
lithology, landform), that are important to the distribution and abundance of ecological
communities in an ecoregion. A 90-m Digital Elevation Model was used in combination
with a bedrock lithology coverage to derive the elevation zone, landforms, and geology
classes used to model ELUs. The final cliff layer for western Maryland was derived from
the Central Appalachian ELU data set.
2.2.3.9.5 Dam/Bridge
This component of the Special Habitat Features layer was originally intended to include
both dams and bridges, but ultimately included only bridges. It was created by
intersecting roads with streams and open water (DLGs and NWI). Although, in many
instances, bridges are not present at stream crossings (i.e., instead there may only be a
small culvert, if the stream is small), this was the only approach available at the time to
create a bridge feature layer for modeling the habitats of bird species that are known to
nest under or on bridge structures, over streams or open water (e.g., peregrine falcon, cliff
swallow, barn swallow). Overpasses and underpasses were also extracted from the
1:100,000-scale Digital Line Graph transportation data set, using minor codes identifying
these features, but these data, which may have improved modeling for certain avian
species (e.g., rock dove), were excluded from the final Dam/Bridge data set. Because of
the problems with this component of the SHF modeling layer, it was not used much in the
modeling.
The first step in developing this component of the SHF modeling layer was to intersect
1:100,000-scale transportation DLG data with 1:100,000-scale hydrography DLG data
and National Wetlands Inventory open water polygons. The intersecting road segments
were then “reselected” into a new line coverage.
2.2.3.9.6 Combining Special Habitat Features
Once all of the individual Special Habitat Feature grids were created, they were either
buffered and converted to grids (e.g., point and line coverages), or they were first
converted to grids and EUCDISTANCE was run in GRID, the results being four separate
38
grids for each feature type, each having a buffer distance (100 m, 2 km, 7 km, 15 km)
considered relevant to a particular species or group of species. The final SHF modeling
layers were created by combining the individual component grids into four binary-coded
"hypergrids," one for each buffer distance, such that the placement of the character in the
binary code denotes the feature type.
Although there are intermediate buffer distances that would be more appropriate for
certain species, an attempt was made to limit the number of grids for simplicity's sake.
Another option that was considered would have involved creating separate modeling
grids for each feature type, and then running EUCDISTANCE just once for each feature
type without specifying an upper limit on distance, allowing for the selection of any
buffer distance based on individual species’ requirements. However, because the original
concept for this SHF layer involved a large number of different feature types, this would
have meant a much larger number of modeling grids to deal with, compared to the final
set of four hypergrids.
2.2.4 Wildlife Habitat Relationships
2.2.4.1 MDN-GAP Species List
The list of species for which wildlife habitat relationships models were developed
includes only those species that regularly breed within the project area. The Delaware
Bay hosts one of the largest concentrations of migrating shorebirds in the Western
Hemisphere (Senner and Howe 1984, Myers et al. 1987), and the wetlands associated
with this bay and the Chesapeake Bay host large concentrations of migrating waterfowl.
Many songbirds and raptors also pass through this region during migration. Various
efforts are currently aimed at conserving the staging areas that support these large
concentrations of migratory birds (e.g., Focus Areas under the Atlantic Coast Joint
Venture of the North American Waterfowl Management Plan, Mid-Winter Waterfowl
Survey, Partners In Flight, Twin Capes program for fall migrations, Western Hemisphere
Shorebird Reserve Network designation of Delaware Bay as a Hemispheric Reserve,
Ramsar designation of Delaware Bay wetlands as Wetlands of International Importance
for migratory birds, National Audubon Society’s designation of the Delaware Bay
shoreline as an Important Bird Area, Shorebird Technical Committee under the Atlantic
States Marine Fisheries Commission, The Nature Conservancy’s Delaware Bayshore
Project, and long-term shorebird population monitoring efforts in both Delaware and New
Jersey). Unfortunately, although MDN-GAP investigators initiated efforts to include
these important staging areas in the Gap Analysis, inadequate project resources prevented
the completion of this component of the project. Therefore, users of the final MDN-GAP
data sets should be aware of this omission, and should consider the results of this project
as complementary to these other efforts when assessing biodiversity conservation
priorities.
Within the three-state project area, there are 41 amphibian species, 47 reptile species, 69
mammal species, and 206 regularly-nesting bird species. These taxonomic groups
combine for a total of 363 animal species for which wildlife habitat relationships models
39
and distribution maps were developed. Regularly-occurring non-native species were
included in this total.
2.2.4.2 Development of Wildlife Habitat Relationships Models
Development of the Wildlife Habitat Relationships Models (WHRM) began with a
compilation of habitat requirements information from available literature. A list of the
most frequently referenced sources is provided in Appendix F. In addition to these
sources, many species-specific studies were also utilized. A summary document of
habitat requirements was created for each species, and that document was then referred to
in filling out a standard form which was used for ranking each of the 103 habitats, in
terms of suitability (unsuitable, marginal, suitable, highly suitable, optimal) for the
particular species, as well as for providing numerical summaries of relationships with
other modeling variables (e.g., relationship to wetlands, elevation, slope, aspect, special
habitat features, etc.). A sample of one of the forms developed for the compilation of
habitat requirements, the one used for birds, is shown in Appendix G. Separate forms
were developed for each taxonomic group (birds, mammals, reptiles, amphibians).
Habitats were given suitability rankings from 1 to 4, with marginal habitats being
assigned a value of 1, suitable habitats a value of 2, highly suitable (or preferred) habitats
a value of 3, and optimal habitats assigned a value of 4. In determining habitat suitability
based on associations described in the literature, terms such as “uses” or “is found in”
were interpreted as indicating that a habitat is “suitable” (value = 2). Terms such as
“favors” or “prefers” were interpreted as indicating that a habitat is “highly suitable”
(value = 3). Terms such as “occasionally uses” were interpreted as indicating “marginal”
habitat (value = 1). The value of 4 was reserved for rare cases where a habitat was
considered “optimal.” In many cases, a suitability ranking may have been based more on
the number of times that a habitat association was mentioned in the literature. If a
particular habitat was not specifically mentioned or inferred through habitat descriptions,
the suitability of that habitat was determined based on the shared characteristics of
habitats that were described.
Once the habitat summary form was filled out, the numerical rankings and weightings
were entered into the wildlife habitat relationships tables. These tables, and the range
data tables, were stored in a Structured Query Language (SQL) relational database. For
each taxonomic group (birds, mammals, reptiles, amphibians), a separate table was
created for each of the modeling variables described in section 2.2.3. A modeling control
table was also created for each group. This table controlled which modeling variables
were used for each species, and the relative weight of each variable. The database was
initially developed in Oracle v. 8.03, and was subsequently exported to Microsoft Access.
It is currently maintained in MS Access 2002. The database tables which were used in
the species habitat and distribution modeling are listed in Table 2.6.
40
Table 2.6. Database Tables Used in Modeling Species Habitat Relationships and
Distritbutions
RANGE TABLES DESCRIPTION
RAN_CONT Controls which of three range mapping approaches is used:
1) BRC data, 2) EST (estimated) range, with added
hexagons, 3) QUAD data (primarily for Rare, Threatened, or
Endangered species)
AM_HEX, AV_HEX,
MA_HEX, RE_HEX
For each taxonomic group, this table controls which
hexagons are included in species’ ranges, based on the
Biodiversity Research Consortium (BRC) data set
AM_RAN, AV_RAN,
MA_RAN, RE_RAN
Table controlling which hexagons are included in species’
estimated (EST) ranges (BRC hexagons plus other hexagons
added based on expert review)
AM_QUAD, AV_QUAD,
MA_QUAD, RE_QUAD
Table controlling which 7.5-minute quadrangles are
included in species’ ranges (primarily for Rare, Threatened,
or Endangered species)
HABITAT
RELATIONSHIPS
TABLES
DESCRIPTION
AM_CONT, AV_CONT,
MA_CONT, RE_CONT
Table controlling which modeling variables (e.g., habitat
type, wetland buffer, aspect) are included in each species’
model, and also includes relative weightings for each
variable
AM_EQ, AV_EQ,
MA_EQ, RE_EQ
For each taxonomic group, this table stores the modeling
equation for each species; modeling equations are similar to
those used in Habitat Suitability Index (HSI) modeling
AM_HT, AV_HT,
MA_HT, RE_HT
Table containing species-Habitat Type (e.g., Oak-Hickory
Forest, Brackish Tidal Marsh) relationships data (i.e.,
suitab