USFWS Gulf of Maine Watershed Habitat Analysis

Arnold Banner, Ph.D. and Sue Schaller
Gulf of Maine Program, Falmouth, ME
March 2001

Abstract: The USFWS Gulf of Maine Program mapped habitats of 64 federally listed Endangered and Threatened species, declining neo-tropical migrants, shorebirds, waterfowl, anadromous and inter-jurisdictional fishes, throughout the United States portion of the Gulf of Maine watershed.  The resulting digital maps can be displayed for individual species, or as combined data highlighting localities having high species richness and high aggregate habitat value. These products may be of use to biologists and managers in Field and Regional offices, state agencies, and non-government entities to evaluate impacts from development activities, to assess wildlife resources at a landscape level, and to focus conservation activities.

Introduction

Conservation planning depends on information on the distribution of species, their habitats, and the ecological factors which affect them. While site inspection gives a most precise view of local conditions, at the time of the visit, biologists generally may benefit from information disclosing longer term conditions and having a landscape or ecosystem perspective. Project biologists and managers do not routinely have time or resources to assemble such background data. This watershed-wide study is intended to provide a comprehensive analysis, narrative descriptions, and display of habitats based on environmental characteristics and available occurrence information, including that from the scientific literature and from unpublished surveys. Habitat maps for individual species have been further processed into composite maps highlighting areas of highest resource value. This type of information has proven useful for screening permits in the Southeast (USFWS Region 4), for refuge comprehensive planning in the Northeast (USFWS Region 5), and for land acquisition with state and non-government partners.

Methods

Selection of Species to Map.  A major aspect of the watershed study is the selection of the species for habitat mapping. Our list of evaluation species was created in a series of steps, starting with a comprehensive survey of species of high national importance occurring within USFWS Region 5. This was developed by combining lists of all federally listed Threatened and Endangered species, 'non-game birds of Management Concern', and waterfowl, shorebirds, anadromous and inter-jurisdictional fishes (inshore species of concern to USFWS and NOAA) which have significantly and persistently declined in abundance. In each case we adopted lists of such species developed by experts/agencies focusing on those groups (see Regional components).

From this overall regional list we extracted those species for which the Gulf of Maine (GOM) watershed offers habitat (based on occurrences in the watershed of those species as shown by Breeding Bird Survey and Christmas Bird Counts, and/or evidence from technical literature, or expert advice, identifying distribution of the species). All of the nationally important species which regularly occur in the watershed were included in our study.

To accommodate concerns of conservation partners with more localized perspective, we created a supplementary species list for the watershed, consisting of species which are both federal trust responsibilities (migratory birds, migratory fishes) and which are designated as Threatened or Endangered by 2 or all of the 3 states in the Gulf of Maine watersheds (see Supplement to the Regional List).

The species for the analysis, then, consisted of all plants and animals from the regional and supplementary lists and which have a relatively significant presence in the study area (more than accidental/occasional occurrences); see the 'SPECIES' column in the table GOM WATERSHED SPECIES LIST.

Study Area Boundary. The GOM watershed boundary was constructed by selecting the outer boundaries of all smaller watersheds in Massachusetts, New Hampshire, and Maine which flow into the Gulf of Maine. These were identified using USGS 1:24,000 and 1:100,000 hydrology coverages. The boundary was extended into the Gulf at Cape Cod and eastern Maine (see figure, below).

Use of GIS for habitat mapping and biological decision making. Geographic information systems (GIS) - are well adapted to handle large geographic databases, to identify and display associations between base themes and species occurrences, and to conduct analyses based on combinations of base layers and/or occurrence information. The digital files may be in the form of grid-cell, point, line, or polygon features and linked databases, which can be displayed and plotted as maps. These maps may represent areas of relatively high habitat value, or areas which we have relatively great confidence are suitable habitat for the species of interest.

GIS often is used just to map and display locations at which geographic features or organisms were observed. Inferences about habitat may be made from such occurrences, particularly by using them with coincident environmental themes, such as land cover or soils types. Alternatively, GIS can be used to create new themes, based on features or combinations of features from one or more existing themes. If these features represent the environmental preferences or tolerances of a species, the resulting theme may depict the local habitat distribution for that species. Such a habitat map can be tested for accuracy or relevancy by comparing the mapped habitats to the locations known to be used by the species. The GIS characterization of habitats based on a species' environmental preferences is a type of habitat modeling.

Approaches to Habitat mapping. Habitats are the places where organisms live (Odum 1959). Therefore, habitats may be identified directly by observing the use of a locality by a species, or indirectly by habitat modeling. Habitats vary in quality, and this can affect the number of individuals which can be supported per unit area, or the frequency or likelihood of their occurrence at a site. We used the observed level of use and the apparent relative suitability of environmental conditions to assign numerical scores reflecting gradations in habitat value. Such scores reflect our level of confidence that an area represents suitable conditions for the species. It must be noted that mapped habitats, whether developed from occurrence information or from models, are necessarily based on data representing interpretations of past conditions.  Even an accurate depiction of suitable habitats cannot guarantee that the species will regularly be found at those sites.  Occupancy of many habitats is highly seasonal and also may be associated with variations in population levels. Our use of models and occurrence data in mapping habitat for each species is listed in the column "basis of final map" in the table GOM WATERSHED SPECIES LIST.

Use of Occurrence Data.  For species subject to recent, comprehensive biological surveys (e.g., aerial surveys of bald eagle nests) we mapped habitats based on these occurrences and scored these areas as 'optimal'. Otherwise, where a survey was less extensive (e.g., International Shorebird Survey sites), we mapped habitats using a model, and used the occurrence information to score as 'optimal' the localities known to be used, placing our highest confidence in the value of those habitats. Where surveys characterized only a subset of the potential habitat (e.g., Breeding Bird Survey stops for many of the species) we used these observations just to test habitat maps derived from models. The primary occurrence information available to us is listed in the table "Occurrence Data Used in Developing Gom Watershed Habitat Maps, or for Testing Modeled Habitats".

Use of Habitat Models.  We developed simple habitat models, similar to the Fish and Wildlife Service habitat suitability index models (USFWS 1980), for use within our GIS. Habitat suitability index models are hypotheses correlating environmental conditions to some measure of a species' biological 'performance', such as abundance or reproductive success. Because of the limited information on those measures for most of our species, our models generally used the probability of occurrence as a measure of performance. That is, our assignment of higher habitat suitability index values represents increased confidence that the species will make use of the site.

Model development included review of the literature and discussions with experts to: 1) identify environmental factors affecting potential use of the study area, and 2) estimate the relative suitability of such habitat features as land cover types, minimum patch size, edge effects, water depths, or soil types. The suitability of each factor was expressed as an index, ranging from 0 (least suitable) to 1.0 (most suitable, in this locality). These individual suitability index (s.i.) values were combined, into an overall habitat suitability index (HSI) value. For example, a preferred substrate type might be accorded an s.i. of 1.0. However, if this was coincident with a totally unsuitable vegetation cover (s.i. = 0), and use of the area was limited by the minimum of these 2 factors, the overall HSI would be 0. HSI may be based upon one to many factors and may consider requirements of various life stages.

Models were used to improve the accuracy of some data sets. For example, the Maine Department of Inland Fisheries and Wildlife (MDIFW) mapped shorebird habitats as polygon features broadly encompassing the areas being observed, with a numerical database listing the abundance of birds at each observation site. Therefore, the polygons surround used habitat and often have inclusions of non-habitat areas. Using a geographic information system (GIS), we applied supplementary environmental data (substrate, land cover, water depth) in order to delete some of the unsuitable portions of these areas.

Model Validation and Adjustment. Draft models for many of the species developed and later were tested using different sets of occurrence data. For testing, we compared the proportion of known occurrences coinciding with mapped habitat to the proportion of a set of random points coinciding with mapped habitat. Where the agreement with occurrences was significantly different (Chi-square; .01 criterion) from the agreement with random points, we regarded the mapped habitats as being a useful predictor. For some species, where the literature on habitat needs was rudimentary, or focused on factors not distinguishable in our environmental data sets, we developed several models and retained the one having the best association with known occurrences. This test encouraged development of parsimonious yet accurate models, since a high level of agreement tended to be prevented by errors of omission (habitat not covering the occurrences) or of commission (habitat too extensive and so covering more random points).

Habitats mapping accuracy is controlled by the quality of the models and of the environmental themes. Narrative descriptions of habitat needs and the resulting species models, as well as metadata for the GIS coverages, is accessible in the Table of Models and Metadata. This table also links to descriptions of some of the major environmental themes used in portraying habitats.

Results

Data analysis was performed using ArcInfo and ArcView software.  Data were processed as shape files and grids (proprietary vector raster spatial data formats, respectively).  Habitat components were developed in successive iterations, and final products were in grid formats.

GROUPINGS OF INDIVIDUAL SPECIES.  Software limitations prevented our combining all species data into one grid. Therefore, species grids were grouped into five general categories. These were: songbirds (grid name [SONGBIRDS]), shorebirds ([SHOREBIRDS]), waterbirds (including waterfowl, seabirds, etc.; grid name [WATERBIRDS]), Federally listed threatened or endangered species ([FED_LISTED]), and the remaining species, mostly fishes and hawks ([FISH_RAPTORS]).  The habitat scores for each of the evaluation species were retained as attributes of the five resulting grouped grids, allowing the user to display maps by species. Alternatively, 'clicking' on a cell when one or more of the grouped grids are active, reveals the scores in that cell for each of those species.  We have developed an ArcView script to allow tabulation of habitat values for all species within a parcel of interest, and comparison of data for a parcel with average habitat values throughout the watershed. The species groupings and associated field names are listed in the Table of Models and Metadata.

SUMS OF ALL HABITAT SCORES. Habitat suitability scores from grids of the individual species also were added on a cell by cell basis to allow users to display the landscape, shaded by overall habitat value.  The scores of this grid (GOMSUM) reflect both the number of species using each cell, and the relative habitat suitability for those species.

HABITAT SCORING BY GENERAL LAND COVER TYPE.  The grid [GOMSUM] was subdivided according to four generalized cover types (grass/shrub/bare, forested, freshwater, and saline). The top increments of habitat values (top 25% and 50%) then were calculated for each of the four resulting grids, considering both the summed habitat scores and total areas of the grids at each score.  This habitat quality times area product is equivalent to the habitat unit values of the Habitat Evaluation Procedures (USFWS 1980), and is intended as a measure of carrying capacity.  Thus, one acre of habitat with a score of 50 should offer the same value as five acres with a score of 10. Protection of the highest value habitat, therefore, confers maximum conservation benefit and efficiency.  The top 25 and 50% increments of the four types were reassembled into the grids [GOMTOPQUARTER] and [GOMTOPHALF], respectively. These grids are useful for viewing the landscape when there is a preference for certain cover types or the user wishes to isolate the more highly valued areas.

Limitations of the Data. As mentioned above, maps of habitats for the individual species are limited by the accuracy of the data sets used in developing them (both occurrence data and environmental themes) and the validity of the models used to interpret those data.

Maps showing some highest increment of habitats (e.g.,[GOMTOPQUARTER]), while providing a useful visual display, may totally omit habitats or habitat components for some species which do not coincide with those of other species, and so never reach a high 'summed score'. Such maps also may leave out lower-scored areas needed to support or buffer the highly scored habitats which are retained.

Artifacts of grid-cell mapping of riverine themes may in some cases cause habitat omissions, and in others aquatic habitat values in primarily upland areas. Fish habitats were gridded from continuous polygons/arcs derived from USGS hydrology coverages. Single line arcs were necessarily converted into strings of cells, each with a minimum width of 30 m (the cell dimensions).  Even where the dominant land cover of a cell was upland, we retained the habitat value for a stream passing through the cell.  As a result, some upland areas will display habitat value for anadromous fishes.  

The grid process left discontinuities when converting narrow (< 30 m wide) polygon features, such as small rivers, where only part of a cell was crossed by aquatic habitat. Moreover, where National Wetlands Inventory polygons representing wetlands and water bodies differed from those of USGS, the only fish habitat we retained was that which corresponded to aquatic classes in the former. Therefore, fish habitats and migratory pathways may not appear as continuous extents of habitat, or extend to the boundaries of contiguous wetlands in all cases.

Suggestions for Updates. We recommend that this project be periodically updated and enhanced. This includes improving the format for ease and accuracy of use and interpretation, and accuracy of the habitat maps. The latter are limited both by our knowledge of the habitat needs of the species, and the quality of the environmental themes. Revisions and corrections should be based on further information from the biological literature, from field checking of the maps and from use of newer and more accurate species survey and environmental themes, particularly land cover and soils.

Acknowledgments. We wish to acknowledge the assistance and information provided by the following individuals and organizations, without which this project could not have been completed.

Robert Barlow (SUNY, NY)
Frank Biasi (The Nature Conservancy, MA)
John Bidwell, Joe McKeon, Graham Smith, Jed Wright, Jean Fujikawa, Robert Houston, Susi vonOettingen, Doug Forsell, Drew Major, Mike Amaral, Sylvia Schmidt, Ron Joseph, Anne Hecht, Janith Taylor, Jennifer Casey, Ann Roy, Kim Sprankle (USFWS)
Mark L. Botton (Fordham Univ., NY)
Randall Boone (Colorado State University)
Ken Buja (NMFS-SEA)
Katie Callahan, Bob Estabrook (NH DES)
Dave Capen, Sean McFaden (University of Vermont)
Dan Coker (Maine Natural Resources Information and Mapping Center)
Diane Deluca, Chris Martin (NH Audubon Society)
Dave Evers (BioDiversity Research Institute, ME)
Don Field (NOAA, CCAP)
Lew Flagg, Seth Barker, Peter Thayer (ME DMR)
Sue Gawler, Molly Docherty (Maine Natural Areas Program)
John Greenwood, Sara Cairns, Ed Robinson, Claire McBane (NHF&G);
Peter Grose (NOAA NOS)
William Hansen (Florida Power and Light, ME)
Brian Harrington (Manomet Bird Observatory)
Karsten Hartel (Harvard University)
Charlie Hodgson (Friends of Taunton Bay, ME)
John B. Holt, Jr. ( MA)
Greg Horton, Norm Dube (Atlantic Salmon Commission)
Bill Huber (Quantitative Decisions, Merion Station, PA)
Keith Johnston, Linda Alverson, Susan Hitchcox (Maine Audubon Society)
Barry Keim (University of New Hampshire)
Harley Knebel, Stephen M. Howard, Keith Pardieck, BH Powell, Dennis Jorde, Rich Signell, Ian Thomas (USGS)
William Krohn, Jeffrey Hepinstall, Raymond O'Connor, Ken Oliviera, Andrew Gilbert (University of Maine)
Marc Loiselle (MGS, ME)
Maine Office of GIS
Massachusetts Office of GIS
Jim McDougal (Essex County Greenway, MA)
Scott Melvin, Brad Blodget, H. Heusemann, Steve McCrae, Amy Maher (MA Fish and Wildlife);
Isabel Morin (ENSR, Acton, MA)
New Hampshire GIS (GRANIT)
Murray Olmstead (MA)
Yvette Ortega (USDAFS)
Dave Sczeback (MA Natural Heritage Program)
Greg Shriver (SUNY, NY)
Carl N. Shuster, Jr. (VA)
Jackie Sones, Kathy Tuxbury (Wellfleet Bay Audubon Sanctuary, MA)
Allen Starr, Charley Todd, Tom Hodgman, John Kenney, Andrew Weik (MDIFW, ME)
Kyle Stockwell, Nancy Sferra (TNC, ME)
University of Southern Maine Library and Interlibrary Loan Department
Peter Vickery (Massachusetts Audubon Society)
Dave Wilkinson (NRCS, ME)
Holly Yachmetz (NMFS, MA)

Citations

Odum, E.P. 1959. Fundamentals of Ecology. W.B. Saunders Co. 546 pp.

USFWS. 1980. Habitat Evaluation Procedures (HEP). U.S.D.I. Fish and Wildlife Service. Division of Ecological Services. ESM 102.