- Shopping Bag ( 0 items )
Gap Analysis: Assessing Protection Needs
J. Michael Scott, Blair Csuti, and Steven Caicco
THE THREAT OF mass extinctions as a consequence of global warming (Peters 1988) joins a growing list of assaults on planetary biodiversity. Other effects of overpopulation (habitat loss, pollution, overexploitation) have driven thousands of species toward the brink of extinction. The U.S. Fish and Wildlife Service currently lists over 1,000 endangered and threatened species, with several thousand more "candidates" on the waiting list. But this is just the tip of the iceberg. Recent estimates (Erwin 1988; Wilson 1988) indicate there are more than 30 million species on earth, but a quarter of them may not see the year 2010 (Norton 1988). Most are insects, many not yet described, that play critical roles in the function of natural ecosystems upon whose biogeochemical cycles we depend for indispensable ecosystem support services (Wilson 1987). The rescue of the current list of endangered species has proved difficult, risky, and expensive. What does this record foreshadow as tens of thousands more species become endangered over the coming years?
Clearly, it is inefficient to save selected species while allowing the natural communities and ecosystems that support them (along with myriad inconspicuous species) to deteriorate. It would be wiser, surely, to identify and manage functioning representatives of each ecosystem type for the maintenance of national biodiversity. While very localized species, likely to be missed by a network of biodiversity management areas (BMAs), would still require individual protection programs, such an integrated conservation strategy would ensure that the vast majority of species never become endangered. Further, if we were certain that the nation's biological resources were secure, noncritical lands could be put to sustainable uses without imperiling native species and ecosystems. The challenge, then, is to build a geographic database of biodiversity to assess the effectiveness of present and proposed BMAs. This process of identifying unprotected species or communities has been described as "gap analysis" (Burley 1988; Scott et al. 1988). Our objectives here are to present the rationale for this new approach for stemming the tidal wave of extinctions threatening the world's natural heritage and to describe the methods used in the model programs in Idaho and Oregon.
DESCRIBING THE DISTRIBUTION OF BIODIVERSITY
The distribution of the resources we are addressing (plants, animals, vegetation cover) is often incompletely documented—a shortcoming that frustrates attempts to develop a national conservation strategy and has led to calls for a national biological survey (Kosztarab 1984; Wilson 1985). But the identification and classification of the nation's insect fauna alone would take years, during which time conservation opportunities would be foreclosed. Nor is there time or funding for the intensive field surveys necessary to document species distributions on the basis of verified specimen records.
The intensive collecting carried out by the Bureau of Biological Survey (predecessor of the U.S. Fish and Wildlife Service) in the late nineteenth and early twentieth centuries forms much of the basis for our current understanding of plant and animal distribution. Despite decades of fieldwork, the ranges of most species are still predicted from the locations of a few specimens and the distribution of habitat types for that species. For example, Bailey (1936:368) suggests that Preble's shrew (Sorex prebeli) "probably has a somewhat interrupted range in the mountains and high country of eastern Oregon" on the basis of the capture locations of three specimens.
Using the association between species and habitat types to predict distribution has a long history of successful application to many common species (Armstrong 1972; Hoffmeister 1986). As Baker (1956:122) succinctly put it: "Mammals generally are confined to specific kinds of plant associations from which they derive either food or shelter or both. Once the investigator has learned the ecological preferences of a given kind of mammal, he can map the occurrence of that mammal by noting the occurrence of the plants." While habitat specificity varies between species and classes of vertebrates, habitat is a powerful predictor of the distribution of many smaller mammals and birds, as well as reptiles and some amphibians.
Direct large-scale documentation of the distribution of an entire fauna is rarely attempted. A recent survey of Hawaiian forest birds (Scott et al. 1986) spanned seven years and recorded a quarter million observations at 10,000 field locations. Application of these methods to the continental United States is clearly impractical. While very rare or localized species are best dealt with on a locality-by-locality basis, we believe that, within known range boundaries, the fine-scale distribution of most species can be predicted from a knowledge of habitat preferences and the distribution of vegetation types.
While an intensive and thorough biological survey is a worthy undertaking, there is a growing consensus that a timely conservation strategy will have to be based on indirect indicators of biodiversity (Roberts 1988). Vegetation, vertebrates, and butterflies are the groups whose distribution is best documented (Pyle 1982; Scott et al. 1987). An analysis of their distribution relative to current nature reserves, national parks, wilderness areas, and the like would yield an assessment of the current level of protection of national biodiversity (Scott et al. 1987, 1988, 1989) and suggest management alternatives to fill the gaps in the network of reserves and BMAs.
VEGETATION AS WILDLIFE HABITAT
Vegetation is the most widely used indirect indicator of biodiversity (Crumpacker et al. 1988). Since it forms the basis for predicting animal distributions, an accurate vegetation map is the first priority of a gap analysis. Kuchler (1964; 1988) has led the way in mapping the distribution of "potential natural vegetation" (PNV) in the United States. The vegetation types depicted on Kuchler's 1964 map of "potential natural vegetation of the coterminous United States" (scale 1:3,168,000) "provide the only assessment of major, aboveground, terrestrial, and wetland ecosystem diversity that describes the entire United States in reasonable detail" (Crumpacker et al. 1988). In these places PNV is a valuable surrogate for biodiversity. Significant areas, however, have been converted to other cover types (such as cropland and pasture, urban and industrial areas, and early successional stages of forests). Wildlife responds to "real" or current vegetation rather than PNV Thus it is possible to use county-of-occurrence information and a current vegetation map to predict the range of a species. Because vegetation forms the basis for predicting range maps, the first step in gap analysis is to create a vegetation map. In the pilot project in Idaho, we used a minimal mapping unit of 640 acres and a scale of 1:500,000.
METHODS FOR MAPPING VEGETATION
Because our time and money were limited, we chose to compile the map of Idaho vegetation primarily from existing sources. About two-thirds of Idaho is public land under the administration of the federal government. Most of this land is managed either by the USDA Forest Service (USFS) or the Bureau of Land Management (BLM). For most of these areas, maps of local vegetation were available from these agencies.
The most recent maps were those produced by the BLM during the past decade as part of the environmental assessment of their management actions. Maps from planning documents were reduced to our working scale of 1:500,000 and then traced onto mylar. In order to maintain a consistent level of discrimination throughout the areas mapped, it was often necessary to generalize the information from source maps of greater detail.
Because the Forest Service's current management practices emphasize site potential, comprehensive information on the actual vegetative cover of lands under its jurisdiction was not available. For these areas, we used timber type maps at a scale of 1:31,680 from the period 1950-1970. These maps were produced through interpretation of aerial photographs. In most cases, types were named on the basis of a single dominant tree species (more than 50 percent canopy coverage); in a few cases, codominant tree species were indicated.
The large scale of these maps, compared to our working scale, made it necessary to compress a great deal of information. This was accomplished visually. Contiguous maps were laid out by township, timber types were color-coded where necessary, and the major forest types within the township were delineated. This delineation was then transferred to mylar at our working scale. There are numerous areas of Idaho for which there are no local vegetation maps. Most of these areas. are in agricultural uses and were mapped as cropland or pastureland. The boundaries of these areas were delineated using a variety of sources, including soil surveys and topographic and geological maps. In nonagricultural areas adjacent to federal lands for which vegetation maps were available, type boundaries were extrapolated. A final edit of the map for agricultural boundaries and areas of recent timber harvest was performed using false-color infrared paper prints of Landsat MSS imagery.
As a result of our work in Idaho, we have modified the mapping procedures for vegetation. Habitat mapping can be done using digital image classification or manual interpretation of satellite imagery.
Both habitat mapping methods have advantages and disadvantages. With manual mapping, the large polygons drawn by interpreters may conform better to habitat classification systems. Moreover, they cost less to store and analyze than digital mapping and, in addition, photo-interpretation requires less training and technology. On the other hand, digital classification can depict greater spatial detail while avoiding errors in boundary placement and digitizing. Digital image classification also retains information on the spatial diversity occurring in habitat mosaics, whereas this diversity is usually generalized to one or another habitat type by photointerpreters.
CHOOSING A MAPPING STRATEGY
The choice of mapping strategy may well be dictated by time and budgetary constraints and is also somewhat region-specific. This, plus the fact that some states may want to produce habitat maps for uses other than gap analysis, makes it impossible to set uniform mapping procedures. In our experience, manual interpretation of specially preprocessed thematic mapping (TM) prints provides an accurate product in vector format that is easily merged with other variables such as landownership and land management. (These linear features are not well suited to the raster data structure of digital satellite data.) Much detail is lost using this approach, but the resulting map polygons are likely to be coherent landscapes well suited to regional analysis and planning. Furthermore, the manual maps are useful in segmenting digital satellite data for more detailed local analyses using digital classification.
The use of different mapping strategies by adjacent states raises the problem of edge- matching maps for broad regional analyses. In our experience, even edge-matching maps produced by the same method and classification system, but using different satellite scenes, can be difficult. The problem of edge-matching maps produced by digital and manual methods is clearly formidable.
PREDICTING ANIMAL DISTRIBUTIONS
The boundaries of traditional range maps enclose known records of occurrence. Unexplored regions may be omitted in error and, due to the small scale of most range maps, areas of inappropriate habitat types may be included. Many conservation and land-use planning decisions require more detailed information on species distributions. Because our 1:500,000 scale map of Idaho depicts polygons as small as 259 hectares (640 acres)—the average polygon size is 11,285 hectares for 118 actual vegetation types—a much finer resolution of preferred habitat types is possible. We used existing descriptions of habitat relationships to predict the presence or absence of native terrestrial vertebrates (with the exception of a few microhabitat specialists) in each vegetation cover type. Since the landscape-scale vegetation polygons include an unpredictable dispersion of seral stages and microhabitat features, we predict a species present within a polygon but not at particular locations within the polygon. And since the timing, degree of dispersal, and habitat preferences of wintering birds are difficult to predict, we considered only breeding bird ranges. Because of the complexity and scale of the vegetation map and the large number of vertebrate species, we are using an ARC/INFO Geographic Information System (GIS) to record and analyze these geo- based "themes."
Current small-scale Idaho distribution maps are available only for reptiles and amphibians (Nussbaum et al. 1983). These maps define the area within which a species is likely to occur in proper habitat. Very small scale distribution maps are available for mammals (Hall 1981), but since they depict continental distribution there are few records from Idaho. The distribution of Idaho birds by 1 × 1 degree blocks of latitude and longitude ("latilongs") is also available. A final guide to establishing range limits is a statewide county-of-occurrence database for all vertebrate species developed by the Idaho Natural Heritage program.
We use the GIS to superimpose two data sets for each species: distribution by county and association with vegetation cover type. The intersection of these data sets becomes the predicted distribution for a species. Initial comparisons with maps drawn by conventional methods are favorable. Each map is reviewed by local experts and further validated by comparison with museum specimen localities. Field validation of the distribution maps is planned for Idaho in 1992.
The set of distribution maps, the vegetation map, a map of biodiversity management areas, and a map of classes of landownership form the data set for the gap analysis. Since the ranges of rare plants and animals cannot be accurately predicted by the distribution of vegetation cover types, specific localities at which these are known to occur are available from the Idaho Natural Heritage program and are included in the analysis as a separate data layer. Biodiversity management areas are defined as areas capable of being managed for the maintenance of native species and vegetation. Since management activities vary for different designations and ownerships, we qualify BMAs with a scale developed by The Nature Conservancy (TNC):
LEVEL 1: total protection of native communities (national parks, TNC preserves)
LEVEL 2: partial protection of native communities (wilderness areas, BLM areas of critical environmental concern)
LEVEL 3: no protection (most private land, nondesignated public lands)
Although a number of different queries are possible regarding the management status of different animal species and vegetation types, the first result of developing a GIS database of vegetation is the creation of baseline information about the distribution and quantity of natural vegetation and its associated species. Since Landsat imagery is used to determine the current extent of cultural land-cover types (agricultural land, urban areas, forest fragmentation patterns), future updates will allow quantification of land use trends on a statewide or regional scale. Landscape patterns—such as the location and width of corridors of natural vegetation between BMAs—can also be mapped and monitored. We believe that, at a minimum, land use status should be updated at ten-year intervals (and more frequently in rapidly changing landscapes).
Excerpted from Landscape Linkages and Biodiversity by Wendy E. Hudson. Copyright © 1991 Defenders of Wildlife. Excerpted by permission of ISLAND PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.