EXECUTIVE SUMMARY

The New York Gap Analysis Project (NY-GAP) was begun in 1993 as a cooperative effort among the Biological Resources Division of the U.S. Geological Survey (U.S. Fish and Wildlife Service at that time), Cornell University, and the NY State Department of Environmental Conservation (NYSDEC). The objectives of NY-GAP were to: (1) produce databases to describe current land cover, predicted distributions of native species of terrestrial (i.e., non-fish, non-marine) vertebrate species, stewardship responsibilities for conservation and selected public lands, and land management status for use in Geographic Information Systems (GIS) at a scale of 1:100,000; (2) identify land cover types and vertebrate species that currently are not represented or are poorly represented in areas managed or potentially managed for long-term maintenance of biodiversity (i.e., identify conservation gaps); and (3) facilitate cooperative development and use of information so that institutions, agencies, and private land owners may be more effective stewards of the biological resources of New York State (NYS).  NY-GAP is a preliminary step toward the more detailed studies and efforts needed for the long-term conservation of biodiversity in NYS – it is a beginning, not an end. 

This report is a summation of a scientific project. While we endeavor to make it understandable for as general an audience as possible, it reflects the technical complexity of the project it describes. A glossary of terms is provided to aid the reader in understanding this report. For those seeking a detailed understanding of the more technical subjects, the literature cited should be helpful. 

New York is a medium-sized state, approximately 12 million hectares (30 million acres) in size, and ranks twenty-seventh in total area (land area plus water area) among the fifty states. However, in terms of human population, with approximately 20 million inhabitants, NYS consistently ranks among the top three most populous states in the United States, with an estimated population density of 1.7 persons per hectare (0.7 persons per acre). With a population of approximately 8 million people, New York City is the largest city in the state and one of the three largest cities in the United States. 

Virtually all of NYS was glaciated 10,000 to 12,000 years ago, with the biodiversity currently represented across the state having repopulated the area following the retreat of the Wisconsonian glaciers. Approximately 200 years of intensive land use, beginning with settlement by western Europeans, has dramatically affected the natural history of NYS, making a complete characterization and mapping of land cover types and other elements of biodiversity a challenging undertaking. It is against this cultural background that we must evaluate elements of New York's natural heritage. The cultural history of NYS, both before and after settlement by western Europeans, is intricately intertwined with its natural history, with a knowledge of each kind of history contributing to a better understanding of the other. 

Land Cover Classification and Mapping 

A statewide map of 29 land cover types was produced to support habitat modeling for NY-GAP. The map was produced using all or parts of 14 Landsat-5 Thematic Mapper multi-spectral digital images acquired during the spring and summer seasons between 1991and 1993, inclusive. A statistical clustering program was used to produce 240 spectrally homogeneous clusters to which a land cover type label was applied for each cluster based on spectral properties, ancillary data, and local knowledge. The nomenclature and procedures of the National Vegetation Classification System (NVCS) were adopted for labeling land cover types. 

When all clusters of all images were labeled into a land cover type using a variety of analytical methods, each scene was digitally mosaiced with its adjacent image and spatially aggregated to a four-hectare minimum mapping unit. Several alternative strategies were investigated to improve the spatial and taxonomic detail of the statewide map using multiple satellite image acquisitions and ancillary biophysical data. Ultimately, time and resource constraints limited map production techniques to trained image analysts labeling clusters derived from single-date, multi-spectral Landsat data. 

A stratified random sample of 113 field plots was used to assess the accuracy of the NY-GAP map using both conventional and fuzzy accuracy assessment methods, as described in detail in the full report. Through the summer and fall of 1998, field crews visited 9,745 map polygons and assigned a land cover type label based on visual inspection and careful assessment of the dominant cover types in the polygon. Each observed polygon was considered a point sample, which was compared to the predicted land cover type for each of the sampled polygons. Field crews were restricted from entering private land to view land cover composition. If polygons were inaccessible, field crews relied on aerial photo interpretation and local knowledge to determine the appropriate land cover type of the polygon. The degree to which the reliance on aerial photo interpretation for assigning land cover type labels to sampled polygons affected the accuracy of the field data was not investigated. Contingency tables at three taxonomic levels were constructed using the field observations. Overall map accuracy at the Super-alliance (NVCS), Subclass (NVCS), and modified USGS Anderson Level I classification levels were 42%, 57%, and 74%, respectively.  

Application of fuzzy accuracy assessment methods using a sub-sample of the 9,745 polygons resulted in 19% and 23% improvement in overall map accuracy at the subclass and super-alliance classification levels, respectively. Fuzzy functions are calculated using linguistic scores that permit greater flexibility in assigning an acceptable answer to an observed land cover type in comparison to what is predicted. There should be further investigation of the extent to which fuzzy accuracy assessment provides additional information useful for improving the quality of land cover maps developed for conservation applications, when compared to conventional accuracy assessment. 

The low map accuracies obtained at NVCS super-alliance and subclass levels have to be balanced with the spatial distribution and area summaries for each land cover type at both levels of the classification scheme. The impact of sampling field plots in close proximity to transportation routes with limited observations for less heterogeneous areas may have resulted in lower map accuracies than would be expected. Additional studies are required to evaluate the impact of the sampling design used in this project. We believe that the accuracy assessment method used in this case possibly undervalues the quality of the statewide map that was produced. 

The NY-GAP map compares relatively well with other regional-scale land cover mapping efforts with regard to area and accuracy of major land cover types. Consistently low producer accuracies for agricultural land cover types are indicative of the difficulties of mapping these particular types. The spatial and spectral variation due to diverse cropping systems, rotations of row crops with hay and pasturelands, and short- and long-term abandonment of cropland contributes to the spectral confusion encountered. Improvements in mapping accuracy of these particular land cover types, which are highly dynamic in their spectral response, can be expected by developing a temporal sequence of spectral indices throughout the growing season or sequence of growing seasons to capture both intra- and inter-annual cropping system spectral variations. 

Predicted Animal Species Distributions and Species Richness 

NY-GAP created predicted distribution maps for 366 terrestrial vertebrate species, including some native and introduced species (e.g., Ring-necked Pheasant and house mouse). All mapped species, except for 8 non-native, introduced species, were included in the gap analysis process for a total of 358 native vertebrate species evaluated using gap analysis procedures. Accuracy assessment was conducted for as many mapped species as possible, including introduced species. Because marine vertebrates are not included in the gap analysis process, five species of marine turtles were excluded from all analyses. Accuracies for three very rare breeding bird species (Merlin, Black Vulture, Wilson's Phalarope) could not be calculated because of a lack of species occurrence records.

Species richness for native, terrestrial vertebrates was computed for each of 249 EPA hexagons within the state, with each hexagon 640 square kilometers in area. Total vertebrate richness ranged from 119 to 285 species per hexagon with a mean of 242.6 (± 27.6). The distribution of species richness values varies among species groups. Birds and reptiles follow a normal distribution, while distributions of amphibians and mammals are skewed to the left, indicating that most hexagons contain a relatively high number of amphibians and reptiles.

The state of New York exhibits some noticeable patterns in predicted species richness. Amphibians and reptiles exhibit greatest species richness values in the Hudson River Valley, with fewer species in the Adirondack Mountains, which is higher in elevation and likely too cold for these species in the winter. The opposite is true for birds, where richness is high in the Adirondack Region, especially in its outer portion (the Adirondack Transition), with somewhat lower relative richness in the Hudson River Valley area. Birds also show another diverse area in western New York in the transition area from the edge of the Appalachian Plateau to the Great Lakes Plain, where there is an abundance of open fields and grassland areas. Mammal diversity is highest in two regions of the state, the Catskills and the outer portion of the Adirondacks, both heavily forested regions. Diversity for mammals is lowest on Long Island, which is about 50% urban, as well as in the Great Lakes Plain, which is about 50% agriculture and open field.

Species richness for predicted distributions also was calculated by 90-meter grid cell. Richness by 90-meter grid cell follows the same general trends as richness by hexagon. The overall accuracy at the town level averaged for all species was 59%. Accuracy was highest for birds at 85% at the town level, reflecting the detailed coverage of the Breeding Bird Atlas data. Accuracy for mammals was approximately 29% because there were fewer observations for mammals at the town level. Amphibian and reptile accuracy fell between the other two groups, at 67% and 65%, respectively. The same methods were used to calculate accuracy at the ecozone level, with an overall average accuracy of 85%.

Accuracy for 33 species of amphibians was lowest for Jefferson Salamander and the Jefferson Salamander complex at 7.0%, with 93% commission error, and highest for the Eastern Tiger Salamander, with 99.2% accuracy and 0.6% commission error. Omission errors were low for nearly all species, except for the Blue-spotted Salamander at 4.6%. Accuracy at the ecozone level and 7.5' quadrangle level ranged from 9.1 – 100% and 10.3 - 96.8%, respectively, for these amphibian species.

Accuracy for 32 species of reptiles was lowest at 14.1% for the Smooth Green Snake with 85% commission error and highest for the Eastern Massasauga, Eastern Mud Turtle and Northern Diamondback Terrapin with 97 - 99.9% and approximately 1% commission error. Omission errors were low overall, ranging from 0 – 2.14%. Accuracy at the ecozone level and 7.5' quadrangle level ranged from 36.4 – 100% and 18.8 – 99.7%, respectively. Accuracy for some of the common species, such as the Snapping Turtle or Eastern Box Turtle fell in the middle of the range at about 50%.

Accuracy for 231 species of birds ranged from 35.2% for Barn Owl and Pied-billed Grebe (with 63% commission error) to 100% for several different species. In fact, more than half of the breeding birds had accuracies above 90%. Omission errors were generally low, from 0 – 1 %, and a maximum of 6.4% for the Northern Bobwhite. Accuracy at the ecozone level ranged from 9.1 – 100%.

Accuracy for 59 mammal species ranged from 4.2% for the Indiana Bat, with a 95% commission error, to 93% accuracy for White-tailed Deer and Rock Vole, with 31% and 62% commission errors, respectively. Omission errors were low, ranging from 0 – 3.6%. Accuracy at the ecozone level ranged from 27.3 – 100%. Besides the high accuracy for White-tailed Deer (93%) and American Beaver (89%), accuracy values for many of the common mammalian species were quite low. For example, the Eastern Gray Squirrel and Virginia Opossum both had an accuracy of about 8.5%. We believe that the low overall accuracy and high commission errors for mammals are a function of limitations of the species occurrence database. Our experience with NY-GAP emphasizes the need for substantial research on the distributions of smaller, non-game mammals in NYS. Overall, distributions and status for small mammals are among the poorest known of terrestrial vertebrates in the state. 

Land Stewardship 

A digital map of land stewardship was created for the state using data contributed from federal, state and private organizations. GAP uses an ordinal scale of 1 to 4 to identify the relative degree of management for biodiversity protection for each land unit, with Status 1 being the most permanent, comprehensive level of protection, and Status 4 being the lowest level, or unknown. In a cooperative effort, representatives from respective agencies and organizations helped assign status codes to public and private organization lands. 

Nearly 10% of NYS was classified as Status 1 or 2 lands. Most of these lands (86%) are located in the Adirondack Forest Preserve and are managed by NYSDEC. Status 1 and 2 lands are disproportionately located at higher elevations, with 70% of lands above 700 m (2170 ft) elevation and 17% of lands below 700 m elevation classified as Status 1 or 2. Approximately 10% of the state was classified as Status 3 and 81% was classified as Status 4. In contrast to the Status 1 and 2 lands, Status 3 lands are well distributed across NYS. State and federal government lands comprise about 14% of NYS, a modest proportion compared to many western states. 

Analysis Based on Stewardship and Management Status 

Our analysis was based on results from intersecting the GIS coverages of land cover types and predicted vertebrate distributions with the stewardship and management status coverages. Generally, the results revealed that forested cover types are well represented on protected lands (those classified as Status 1 or 2), but nearly all shrub and grasslands are under private management and are poorly protected.

Predicted distributions for a total of 358 native, terrestrial vertebrate species were mapped for NYS. Of these species, only five were not predicted to occur at all in Status 1 or 2 lands (Eastern Hellbender, Shorthead Garter Snake, Ring-billed Gull, Clay-colored Sparrow, and Yellow-throated Warbler). A large proportion (72%) of vertebrate species in the state have less than 10% of their predicted distribution on Status 1 or 2 lands. Nearly all reptile species (97%) and numerous amphibian species (75%) are in this group. A majority of mammal species (64%) and bird species (70%) are also in this group.  

There are 66 species of vertebrates currently listed as being of special concern in NYS (U.S. Fish and Wildlife Service, NYSDEC, and The Nature Conservancy). Among these species, nearly all (92%) of the amphibian and reptile species, 76% of the bird species, and 66% of the mammal species have less than 10% of their predicted statewide distributions within Status 1 or 2 lands. Many species that appear to be least protected are associated with grassland habitats or habitats associated with water bodies. Cover types that represent these habitats are not only uncommon, but also among the least protected areas in the state. There are a few listed species, such as Bicknell’s Thrush, that prefer higher elevation areas in coniferous forest cover. Cover types representing these habitats have some of the highest proportions of occurrence in Status 1 and 2 lands, and it logically follows that nearly 90% of the expected distribution of Bicknell's Thrush is on protected lands. 

Conclusions and Management Implications 

More than 200 hundred years of intensive land use by humans, beginning with the first waves of western European settlers, have created a complex landscape mosaic across NYS. Widely dispersed land cover types of small size and substantial within-type diversity present an ongoing technical challenge for landscape characterization. The vegetation map of NY-GAP represents the first time in history that a digital land cover map, including an accuracy assessment, for the entire state has been created and analyzed, based largely upon satellite imagery. This map, however, represents a beginning, not an end. Only if subsequent maps are produced in the future will we be able to assess change in land cover types and extent across the state. Such land cover change detection is essential, however, if we are going to be able to estimate the spatial nature and temporal trajectories of land cover change and evaluate conservation options in that dynamic landscape context. 

The procedures of gap analysis, developed by professionals, widely tested, and thoroughly peer reviewed, offer a protocol and template for continued collection and use of spatially referenced information about the occurrences of elements of biological diversity at the species level and above. As more management practitioners develop a better understanding of new technologies and become more familiar with applications of remotely sensed imagery, digital image processing, and geographic information systems, greater acceptance and application of these technologies can be expected. Potential applications of GAP data still can be limited by user or institutional inertia, an absence of creativity and imagination on the part of users, and an absence of adequate funding for exploratory studies and refinement of methods. For example, issues related to the propagation of errors in all phases of the gap analysis process still need to be investigated more thoroughly. Errors contained in the land cover map, habitat matrices, vertebrate models, validation data, and land stewardship maps may all contribute to the uncertainty of predictions, relationships of animals to habitats, and map products resulting from the gap analysis process. 

New York is a densely populated state, with only approximately 15% of its land area represented by public lands, actually or potentially managed for biodiversity conservation. This proportion of public lands is in marked contrast to the picture for most of our large, western states. For example, we have learned that substantial observed terrestrial vertebrate biodiversity is represented within the corridor along the Hudson River, between Albany and New York City. The Hudson Valley Region, with 13.5% of the land area of the state and only 12% of the state's public lands, has documented from within its boundaries 83% of the state's terrestrial vertebrate diversity, 86% of the land-cover diversity, and 54% of the state's human population. These statistics capture the essence of the challenges which face biodiversity conservation for our eastern United States in the twenty-first century. 

Information from the GAP database about land cover types and predicted occurrences of terrestrial vertebrates could be useful in guiding a number of state inventory and planning efforts. Among such efforts can be included open space designation, planning, and management, along with development of long-range management plans for state wildlife management areas and state forests, where GAP data can be used to map land cover types and estimate land cover type areas for state lands and also offer a first approximation of expected (i.e., predicted) terrestrial vertebrate species diversity to aid in the planning and inventory process for land management units. 

With completion of NY-GAP, we are poised to address a number of questions relevant to the regional content and context for elements of coarse-filter and fine-filter biodiversity within NY. Typical content questions that can be asked for any study area of interest within NY are of the following form: 

How many and what kinds of vertebrate species and vegetative associations are found in the study area?  

Which ecological or political subunit (e.g., ecoregions, counties, towns) has the most amphibian species? Reptile species? Bird species? Mammal species? Vegetative associations? 

What is the ranking of ecological or political subunits of the study area for elements of biodiversity represented within their boundaries, from most diverse to least diverse? 

What is the ranking of public lands in the study area for the elements of biodiversity represented within their boundaries, at both fine-filter and coarse-filter levels, from most diverse to least diverse? 

Within the study area, which species and vegetative types are well represented on public lands and which are poorly or not at all represented on public lands? 

Likewise, additional questions related to landscape context for a study area of interest could be addressed: 

Are there any terrestrial vertebrate species or vegetative associations found only in the study area and nowhere else within the larger area in which the study area is embedded? 

What proportions of the fine-filter (i.e., species) and coarse-filter (i.e., vegetative associations) biodiversity from the larger region in which the study area is embedded are represented in the study area? 

It is hoped that there will be sufficient interest from agency leaders and others for these kinds of questions to be addressed in the future for a variety of regions of NYS, both ecological and political. The gap analysis process and NY-GAP database provide the foundation and basic information essential for these kinds of analyses and associated planning efforts at landscape scales to be done in the future.

ACKNOWLEDGMENTS

Thanks to Amos Eno and the staff of the National Fish and Wildlife Foundation, who funded the early development of the GAP concept.  Thanks to John Mosesso and Doyle Frederick of the Biological Resource Division (BRD) Office of Inventory and Monitoring for their support of the national GAP program, especially during the transition from the U.S. Fish and Wildlife Service to the National Biological Service and then to the U.S. Geological Survey Biological Resources Division.  Thanks to Reid Goforth and the staff at the BRD Cooperative Research Units for administering GAP’s research phase from headquarters.  Without those mentioned above, there could not have been a Gap Analysis Program.  Thanks also to the National Gap Analysis Program staff.

We acknowledge contributions to this report by Chris Cogan, Patrick Crist, Blair Csuti, Tom Edwards, Michael Jennings, and J. Michael Scott.

Gap analysis is, of necessity, a broadly based, multi-disciplinary, multi-agency effort.  The staff of the New York Gap Analysis Project (NY-GAP) would like to express our sincere appreciation to the following agencies and individuals who have contributed their time and energy toward completion of this important project:  Adirondack Park Agency (Avram Primack), Environmental Protection Agency, Federation of NYS Bird Clubs, NY Natural Heritage Program (Kathy Schneider, Andy Finton, and Greg Edinger), NYS Department of Environmental Conservation (Gerry Rasmussen, Division of Fish, Wildlife, and Marine Resources; Kurt Swartz, Division of Lands and Forests), NY State Museum, NYS Office of Parks, Recreation, and Historic Preservation (Tom Lyons, Robert Reinhardt), The Nature Conservancy (Kate Hubbs, Bill Brown), U.S. Department of Defense (Rich LeClerc, James Beemer), U.S. Forest Service (Clay Grove, Chris Zimmer, Rachel Hershey), U.S. Fish and Wildlife Service (Region 5), and U.S. Geological Survey.

We also would like to thank reviewers who commented on drafts of this report.  The reviewers were Katherine Barnes, Alvin Breisch, Russell Cole, Nick Conrad, Scott Crocoll, Pete Gradoni, Robert Miller, Paul Novak, Gerald Rasmussen, and Laura Sommers of the NYS Department of Environmental Conservation, Kathy Schneider and Greg Edinger of NY Natural Heritage Program, Patrick Sullivan from the Department of Natural Resources at Cornell University, and Patrick Crist from the National Gap Analysis Program in Idaho.

In addition, cooperation and collaboration has occurred among a variety of Cornell departments and units, each providing different expertise to contribute to NY-GAP’s success.  Among those Cornell departments and units are the following:  Center for the Environment, Cornell Cooperative Extension, Cornell Institute for Resource Information Systems (CIRIS), Cornell Theory Center (National Supercomputer Facility), Department of Natural Resources, Department of Crop and Soil Sciences, Soil, Crop, and Atmospheric Sciences, and School of Civil and Environmental Engineering.  In particular, we express our appreciation to Jason Beecher, William P. Brown, Robert Elliot, Anna Stalter, Kim Sweeney, and Saphida Warimu for their contribution to the NY-GAP databases; to Daniel Decker and James Lassoie who provided a “home” for the project and its staff in the Department of Natural Resources, College of Agriculture and Life Sciences, at Cornell; and to Ellen Harris, Administrative Assistant for the NY Cooperative Fish and Wildlife Research Unit.