Identification of Important Fish and Wildlife Habitats
of Southern Maine

Arnold Banner
U.S. Fish and Wildlife Service Gulf of Maine Program
June 12, 1998

Introduction

The Gulf of Maine Program conducted a geographic information system (GIS) analysis to identify and rank habitats important to priority "trust species" of the U.S. Fish and Wildlife Service (USFWS) in southern Maine. Results from this project will be used in establishing a new boundary, and in comprehensive planning for the Rachel Carson National Wildlife Refuge (RCNWR). It also may be of interest for other conservation initiatives. While the Refuge, Wells National Estuarine Research Reserve, the state and private organizations protect valuable areas in southern Maine, the assistance of willing landowners, other federal and state agencies, statewide conservation groups, local land trusts and town planners is needed to protect additional lands.

This web site contains background information on the data and models used to create the maps.   A cdrom is available from the Gulf of Maine Program which contains background information on the data and models used to create the maps as well as a complete set of habitat maps, composite maps highlighting particularly valuable tracts.

Study Area
The study area includes 14 coastal and near-coastal towns in Cumberland and York counties, Maine: Cape Elizabeth, Scarborough, Old Orchard Beach, Saco, Biddeford, Kennebunkport, Kennebunk, Wells, Ogunquit, South Berwick, York, Arundel, Eliot and Kittery (Figure 1). The study area is approximately 128,000 ha (316,000 acres), encompassing current RCNWR coastal holdings, and inland areas already known to retain an impressive diversity of terrestrial species and which are experiencing a high rate of development (McCollough et al. 1996). This analysis covers the geographic area between earlier USFWS habitat mapping projects encompassing Great Bay, New Hampshire (immediately to the south), and Casco Bay, Maine (immediately to the north).  We have supplied coverages of town boundaries and names on the cdrom (in the directories /data/towns, /data/townames) as overlays for referencing the results of the analysis.

Fish and Wildlife Resources
We identified habitats of 43 endangered species, migratory birds (including non-game birds of management concern, shorebirds, and waterfowl), and migratory (inter-jurisdictional and anadromous) fishes. The individual maps were combined to identify areas with high richness and habitat quality for the evaluation species. We also mapped large contiguous areas of undeveloped lands, and existing protected lands in the study area. These are likely to be useful for conservation and management purposes.

Acknowledgments
A major portion of this work was performed ably by Gerry Hayes. Other Gulf of Maine Project and Rachel Carson NWR personnel contributed ideas and advice. Vital data was made available by the Maine Office of GIS, NOAA CCAP program, Maine Geologic Survey, MDIFW, and MDMR. The FWS Migratory Bird Management Office provided important information for selecting the evaluation species.

Methods

Selection of Evaluation Species
The Fish and Wildlife Service is responsible for species which typically migrate across state and national boundaries, as well as federally listed threatened and endangered (T&E) wildlife and plants. For this analysis we selected a subset of the federal 'trust species' - those which have seriously reduced populations on a national, Gulf of Maine watershed, or state of Maine basis. Trust species were included if they were known to occur in the study area more than occasionally, and were:

1) threatened or endangered nationally (federal listing), or

2) listed by 2 of the 3 states within the Gulf of Maine watershed, or

3) listed by Maine, or

4) experiencing persistent long term declines in populations over much of the U.S. range.

Sources used to identify fish and wildlife species in decline included: Non-game Birds of Management Concern in the Northeast, 1995 (USFWS web site); Waterfowl Population Status, 1996 (USFWS web site); Waterfowl Population Status, 1997 (USFWS MBMO, unpub.); International Shorebird Survey Report, 1995 (USFWS web site); Status of the Fishery Resources off the Northeastern United States for 1994 (NOAA 1995).

The evaluation species are listed in Table 1.

Table 1. Southern Maine Species List

Identification of Habitats
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 delineating locations having environmental conditions preferred by the species. Habitats vary in quality, and this can affect the number of organisms 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 relative suitability of environmental conditions to assign numerical scores reflecting gradations in habitat value. These scores may be related to the carrying capacity of an area, and of our level of confidence that it may be occupied. It must be noted that maps based on observations of past occurrences, or on the coincidence of suitable environmental conditions do not guarantee future use of an area by the species.

Use of Occurrence Data. We used biological survey information to identify habitats and also to test certain habitat maps derived from models (see below). Examples of occurrence data are the shorebird and waterfowl concentration areas, provided by Maine Department of Inland Fisheries and Wildlife (MDIFW). These data sets include a numerical database (abundance of wildlife at each observation) and polygon features broadly encompassing the areas being observed. Therefore, the polygons may have inclusions which are not suitable habitat. Using a geographic information system (GIS), we applied supplementary environmental data in order to delete some of the unsuitable portions of these areas (e.g., having inappropriate water depths), or to highlight features which might be the resources attracting wildlife to the area (e.g., shellfish beds as forage for waterfowl).

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 relating the association of environmental conditions to some measure of a species' biological well being, such as abundance or reproductive success. Because of the limited information on those measures for most of our species, our models generally assign higher habitat suitability index values to sites at which we have the most confidence that the species will occur.

For each species, model development included review of the literature and discussions with experts to identify and estimate the relative suitability of such habitat features as landcover types, 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) relative to conditions available in southern Maine. These individual suitability index (s.i.) values were combined, as a product or mean, into an overall habitat suitability index (HSI) value (USFWS 1980). 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, may consider requirements of various life stages, and may based upon assessments at any spatial scale.

The background information and its application to our environmental and occurrence data are described specifically in each of the model narratives. Characteristics of the resulting coverages are summarized in the metadata section at the end of each model narrative. It must be emphasized that these models may only be applicable to habitats in southern Maine; alternative combinations of environmental conditions and behavior of the species may allow use of other resources in different localities. Also, the map products have not been field verified, and so represent only potential habitat.

Biological Data Sets. Habitat maps incorporated data from MDIFW: shorebird roosting and feeding areas, coastal wildlife concentration areas, seabird nesting islands, biological conservation database coverages. We also used the Maine Department of Marine Resources (DMR) shellfish coverage. Other data sets were used when examining, testing, or adjusting models (see below): the USFWS breeding bird survey, woodcock survey, seaside sparrow survey by MDIFW, and bird surveys by RCNWR.

Environmental Data Sets. These models compute habitat suitability according to the correspondence of the type or level of each environmental factor, with the preferred conditions. Thus, the identification of habitat depends on the accuracy of both the models and the environmental base maps to which the models are applied. Environmental data sets used in the southern Maine analysis include bathymetry (derived from NOAA soundings and nautical charts), wetlands (USFWS National Wetlands Inventory digital maps), coastal and surficial geology (Maine Geologic Survey), soils (USDA/NRCS, landcover (NOAA/CCAP), and USGS hydrology and roads (supplied by Maine OGIS). Certain of these themes have been supplied on the cdrom, such as /data/bathy.tif, the vector coverages ~/wetlands, ~/coast, ~/ponds, ~/rivers, ~streams, and ~/roads.

Spatial Data Processing
Mapping was done using a geographic information system (GIS); all data layers were created as or converted to grid cell (raster) format, in UTM zone 19; units in meters, NAD 27. The analysis relies on data collected by aerial survey and remote sensing. We used 15 m square cells, which approximates the minimum mapping resolution of line features from 1:24,000 scale vector coverages (NWI and USGS). The 30 m cell size of the Landsat data was rescaled to 15 m allow insertion of higher resolution data, where available. The species narratives were converted to a sequence of commands in ArcInfo; all steps and intermediate products were documented and archived at the Gulf of Maine Program. Specifics of the GIS products are described for each theme and may be accessed via the table at the end of this report.

Model Validation and Adjustment. The draft models were used to produce habitat maps. Where occurrence data was available (11 species), these maps were compared to the known distribution of the species. Models then were adjusted, insuring that they 1) were in agreement with the literature on the preferences and tolerances of the species; 2) specifically or narrowly defined the suitable conditions to avoid over-estimating the value and availability of habitat, yet 3) displayed suitable habitat in the vicinity of observed occurrences of the species.

Assembly of data layers. Habitats were mapped for all 43 species, 16 of which had multiple coverages (e.g., roosting and feeding; reproduction and wintering). In order to interpret this complex array of data, we also produced a composite coverage which included habitat information for all species. Habitat scores from the individual coverages were added ("overlaid") on a cell by cell basis. In order to give equal weight to each species, where multiple coverages existed we applied the maximum value per cell from all of that species' coverages.

The composite displayed the overall range of habitat values regardless of the underlying land cover type. To display habitat value by cover type (e.g., show the relatively highest value grasslands, or highest value forested areas) we made composites of habitat scores for each of 4 major landcover classes: 1) grass, shrub, and bare land; 2) forest; 3) freshwater aquatic and fresh emergent wetlands; 4) saltwater, estuarine and saline emergent wetlands. Thus, the user can select highly scored examples of one or all cover classes.

Mapping of Large Tracts of Undeveloped Lands
In southern Maine, concern has been expressed over the rate at which natural landcovers are being replaced by residential development. This "suburban sprawl" includes encroachment by a network of paved roads, buildings, and lawns which interrupts movement of terrestrial and aquatic organisms, and favors access of competitors, predators and parasites to otherwise homogeneous habitats (e.g., forest interiors). Fragmentation of landscapes may be adverse to organisms because of isolation of populations, increased vulnerability to predators along new habitat edges, and reduction of habitat area to levels which can no longer support the species. Our mapping of habitats for the evaluation species included consideration of edge effect and patch size. Although most birds are highly mobile, habitat fragmentation may affect species which have high fidelity to specific nesting localities. Mammals, reptiles, amphibians, and some invertebrates are particularly likely to be affected by fragmentation from development activities, and so biological diversity may be conserved by focusing protection on the relatively large tracts of natural lands remaining in the study area.

Large tracts of undeveloped lands were identified in two steps. Using the GIS, we first selected the paved roads in the study area. The roads were used as polygon boundaries, enclosing tracts with natural and developed landcovers. We next calculated the area of natural undisturbed landcovers within each polygon. We included as natural landcovers all classes except those designated "developed" or "agriculture". Sites within 90 m of development were regarded as disturbed or degraded, and so also were excluded. The tracts were sorted according to their area of natural undisturbed land (100 to 500, 500 to 1000, 1000 to 5000, and over 5000 acres). A map of these tracts is presented in the graphics file /data/unfrag.tif on the cdrom. In order to identify highest priority lands for conservation we performed an overlay analysis to select the highest value habitats within relatively large tracts (> 1000 acres).

Mapping of Conservation Lands
Lands which are managed for conservation purposes were identified so that users could see which habitats were protected currently, and which adjacent lands would most effectively supplement wildlife values of those holdings. The acquisition of adjacent lands would allow increased efficiency and diversity of management, and conservation or restoration of larger habitat patches.

Conservation land holdings were identified from several data bases. These included a statewide conservation and protected lands coverage (1:100,000 scale; Maine Office of State Planning), RCNWR parcel boundaries (1:24,000, USFWS Office of Realty and GOMP), large conservation and public tracts in Wells and Kennebunk (1:24,000, Nature Conservancy), York Land Trust holdings (we digitized from parcel maps). We included all areas that were publicly owned or owned by a conservation organization, regardless of the primary management purpose. Thus, the sites include lands which are managed for wildlife (e.g., RCNWR, Scarborough Marsh Wildlife Management Area, Kennebunk Plains), and other tracts owned for water supply purposes (Wells/Kennebunk well field). A draft coverage of conservation lands is supplied on the cdrom (/data/conserv).

Results

Maps were produced for the following life stages of the evaluation species:

Table 2. Stages Mapped by Species

* rep = reproduction; juv = juvenile; feed = feeding; roost = roosting; migr = migration

Models for all of the evaluation species can be accessed through the table at the end of this report.  Maps for all species are in the data/ directory on the cdrom and can be accessed via ArcExplorer.

Examination of the composite coverage (sum of all species maps) showed that the data were skewed (i.e. many cells had low scores, and a relatively few cells had the highest scores (maximum of 184; see graph Figure 2). Indeed, the total area of those habitats having the top half of the scores (i.e., summed scores ranging from 92 to 184) is less than 1% of the total habitat area.

This extreme result was moderated when the results were broken up among the 4 general cover types, and when presented as area weighted values rather than just summed scores (see below). The distribution of species among the four major land cover types is shown in Table 3. Some species are restricted to one of the cover types, but others are listed in more than one column (e.g., northern harriers, which feed over grassland, fresh and salt marshes). It is apparent that use of estuarine and marine habitats was most widespread.

Table 3. Number of Evaluation Species using the Major Cover Types

forested	bare/grass/shrub	fresh emergent/aquatic	salt emergent/aquatic
9	11	8	26

The areal extents of these four cover types within the study area is shown in Table 4. Note, for example, that our study area was one-half forested.

Table 4. Area and Percentage of the Study Area in the Major Cover Types

The distribution of the area-weighted scores (habitat unit values) within these four cover classes is displayed in the following graphs (Figures 3 - 6). Area-weighting acted to normalize the distributions. We calculated habitat unit values by multiplying each habitat suitability level (each unique sum of scores for all species) by its respective number of cells.

These area-weighted scores were summed cumulatively, and we identified the habitat suitability level representing the midpoint of the range of habitat unit values. In theory, half of the overall habitat value lies above and below this score. This approach reduced the influence of the few extreme values, while allowing us to select just medium and high value examples of the four major cover types. Maps of all habitats for each of the general cover types are shown in the graphics files /data/forestsp.tif, ~/grasssp.tif, ~/saltsp.tif, and ~/freshsp.tif on the cdrom.

Habitat Units for each Habitat Suitability Index Value

for the Four General Cover Classes

We examined the relative proportion of the study area occupied by these highest value habitats. The findings are presented in Table 5.

Table 5. Area and Percentage of the Study Area having the top Half of Habitat Unit Values

Comparison of Table 5 to Table 4 shows that half of the habitat value resides in half or less of the area of each of the cover classes. Thus, we have identified the cells offering maximum habitat value per unit area. A map of the top half of habitats, by cover type, can be viewed in the graphics file /data/tophalf.tif on the cdrom.

The final selections were done to insure that we identified tracts meeting minimum area requirements of sensitive species and that were large enough to offer diversity and management potential. This was the overlay analysis, mentioned earlier, involving large tracts of high value habitats. We developed two versions, both restricted to cells having the top half of the habitat unit values. In each case we selected only habitats which were situated within the tracts of natural lands larger than 1000 acres, or were part of aggregations of contiguous cells (of any of the four types) having a minimum area of 100 or 50 ha, respectively. The results are presented in Table 6, and illustrated in the graphics files /data/top100.tif and ~/top50.tif on the cdrom.

Table 6. Area and Percentage of the Study Area having the top Half of Habitat Unit Values
and within Large Tracts

*Within 1000 acre or larger tracts, or part of 100 or 50 ha aggregation

We estimate that only 6% (7,656 ha) of the study area is designated for conservation or wildlife use, and of this only a relatively small proportion (30%) constitutes the highest value habitats. Looked at another way, the following table shows the area and proportion of the high value habitats that are currently protected, by general cover type.

Table 7. Percentage of the High Value Habitat (Large Tracts) in Conservation

Discussion

The analysis identifies areas of habitat value to 43 birds and fishes, and also shows habitats which are part of large, undeveloped tracts. We derived subsets of these areas which have the highest aggregate habitat values, and which offers ecological diversity and conservation potential based on the extent of the tracts.

Table 7 shows that upland cover types, particularly grasslands, are currently under-represented in conservation ownership. This is particularly significant, considering that the relatively higher level of conservation ownership of fresh and saline wetlands is reinforced by existing regulatory protection. That is, there are state, federal and local limitations to development of wetlands (as well as practical engineering considerations), which do not exist for forested and grassland areas.

Information in Table 5 implies that protection of the top half of habitat values would require protection of resource values of about 20 % of the study area. Conservation of the top half of all types that are within large tracts theoretically could be accomplished by protecting a total of 13% to 15% of the study area (Table 6), and 30 % of this goal already is met in existing conservation holdings. In practice, meeting this goal might require a 'rounding up' to insure the inclusion of some features relied on by the individual species models, but which do not overlap the top half of the habitat values. For example, a shorebird's feeding areas may be in the top half of habitats if other species also used those areas, but if its roosting habitat was relatively unique, that component might not score in the top half of the 4 general cover types, and so might not be highlighted as a conservation priority. Therefore, the shorebird foraging habitat would only be usable for as long as roosting habitat was not developed. The requirement for such alternative cover types exists in 13 of the 43 species. By overlaying the most restrictive coverage (Figure 14, top half of the habitats, within or part of tracts > 100 ha) on each habitat component of these 13 species, we found that only coastal island habitats were completely omitted. Several islands in the study area are used for nesting by common, arctic, and roseate terns, and for roosting by killdeer. However, these islands are already in conservation ownership, and so the omission is not of consequence.

We urge the following cautions in use of these data. The analysis represents habitats for a relatively large number of species, but all are migratory birds and anadromous or inter-jurisdictional fishes. The evaluation species were limited to those experiencing population declines or already at low levels (endangered or threatened status). Although these species act as an 'umbrella' for other organisms in their communities, the identified habitats may not meet the needs of indigenous plants, invertebrates, or other vertebrates. Important habitats identified by agencies with other perspectives (e.g., the Nature Conservancy, MDIFW) would be a valuable supplement. Moreover, wildlife populations increase, decrease and change location over time; this analysis is based on the best available information, as of 1998. The analysis identified potential habitats of various levels of quality; these sites have not been field checked to verify the occurrence of the evaluation species. In addition to the need for verifying the species models, we see a major benefit from improving certain environmental and biological data sets. Specifically, these include shellfish and eelgrass distribution, soils, finer distinctions in mapping of forest and grassland cover types, and conservation lands.

The analysis should prove most useful at the landscape level, for identifying patterns of habitat value by species or cover type. Typical applications may include: planning for the Refuge, open space planning for state, county, or local needs, identification of large tracts within which conservation interests may acquire land or protective easements. For applications the size of a few acres or less (e.g., screening permit applications, conservation of smaller parcels) this information may be useful for identifying features which then should be checked by field inspection. The accuracy of the basic environmental data sets can be verified or corrected, and the site specific conditions can be used with the models to yield a locally improved estimate of habitat values.

Literature Cited

McCollough M., S. Gawler, and J. Albright. 1996. Rare and endangered wildlife, plants, and natural communities in York County wetlands. Maine Natural Areas Program. 27pp.

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.

USFWS. [web site]. Non-game Birds of Management Concern in the Northeast - The 1995 List. http://migratorybirds.fws.gov/reports/speccon/tblconts.html

USFWS. [web site]. Waterfowl Population Status,1996
http://www.fws.gov/~r9mbmo/reports/status/coversht.html

USFWS. [web site]. 1995 International Shorebird Survey Report, March
http://www.im.nbs.gov/iss/95iss.html

NOAA. 1995. Status of the Fishery Resources off the Northeastern United States for 1994. Tech. Memo. NMFS-NE-108.

Appendix A

Distribution of Species Among the General Cover Types

Appendix B

Southern Maine Data Themes