Status of GAP Components This section reviews the status of the constituent parts, or components needed, to conduct Gap Analysis. Gap Analysis was begun with a focus on the terrestrial environment, however, the development of information and analyses must logically be extended to the aquatic environment; the aquatic component of Gap Analysis is also treated in this section. The status of analyses of the GAP data layers is discussed in the "Products" section of this report. Vertebrate Species Maps One of the most powerful contributions made by GAP is the development of consolidated digital biogeographic data for each native vertebrate species which are consistent across all states. The value of the land cover data layer is enhanced with the addition of the vertebrate species data layer. Previous to the development of GAP species data sets, there was no available source of meso-scale (i.e., 1:100,000) maps showing individual species distributions across large areas. With the GAP data, a user can display information (in map, tabular, or text formats) of the distributions of any single vertebrate species or any combination of species of interest. Since the purpose of GAP is to provide an objective, spatially explicit basis of information for local, state, and national options in managing biological resources, the vertebrate species maps are a critical component. The GAP data for vertebrate species distributions have so far been used to create draft state wildlife atlases in two states (Oregon and Idaho). All species range maps are predictions about the occurrence of that species within a particular area (Csuti 1994b). Gap Analysis species maps predict the distributions of species at a landscape scale. (A landscape is made up of more than one kind of natural community, generally covering several thousand hectares [Whittaker 1977]). Because the occurrences of most species must be sampled from collections made at individual points, and these point samples are usually compiled into general small-scale maps (e.g., 1:10,000,000) in field guides, landscape-scale information on species distributions is often lacking in land management decisions. In addition to the predicted distributions of vertebrate species, the consolidation and digitization of specimen collection records and literature, which is required in order to construct this data layer, constitutes progress in the informed management of individual species. Much more importantly, though, is the enormous step forward that this development represents in managing assemblages of species. Species which are not threatened with extinction or not managed as game animals are not generally monitored as they occur throughout their range. Thus, their incremental decline because of habitat loss can, and does, result in one threatened or endangered species "surprise" after another. Frequently, the records that do exist for an ordinary species are truncated by state boundaries. Simply creating a consistent spatial framework for storing, retrieving, manipulating, and analyzing the totality of our knowledge about the status of each vertebrate species is one of the most necessary and basic elements for preventing further erosion of biological resources. The procedure used to map species distributions is shown in Figure 3. A detailed description is provided by Csuti (1994b). One of the central issues surrounding the vertebrate distribution maps has been their assessment for accuracy, which is not shown in Figure 3. The development of an approach for assessing the accuracy of the vertebrate maps is discussed in section 1 below. Another advance is the application of the EPA's Environmental Monitoring and Assessment Program (EMAP) hexagon grid system for geographically smoothing the locality records as a part of the model input.
An estimate of the uncertainty in the vertebrate species distribution maps is critical to the appropriate use of GAP information (Kareiva 1993). To address this issue, guidelines for assessing the predictability of the species maps were drafted by a committee of zoologists and biogeographers for the GAP handbook (Cassidy et al. 1994). Three approaches to assessing the maps are provided: (a) independent expert review, (b) using existing inventories and checklists for comparison, and (c) field surveys. The GAP standards require a minimum of (a) and (b) as a part of the deliverable products. Recognized experts on each species are asked to review the GAP map for that species and to rate overall map quality. The reviewers are asked then to provide a statement on their confidence in their evaluation (confident, moderately confident, somewhat confident, best guess). These levels of confidence provide users with insight into potential biases or weaknesses in the data. They also may identify areas that need further study and can target areas for future field sampling in a coordinated manner across many different agencies and organizations. Finally, in addition to estimating overall map quality, the individual components used to develop the data are assigned levels of confidence, most importantly, the species-habitat matrices, the overall distributional limits, and the quality of the land cover map.
Figure 3. The steps used for initial development of vertebrate range maps for Gap Analysis (after Csuti 1994b). The second requirement for assessing the predictability of the species maps is to compare them with field-generated comprehensive, recent, and reliable checklists of species for the same area. While this method depends on the availability of such lists, the GAP experience has been that they are available for sizable areas such as wildlife refuges, national parks, national forests, and industrial forest lands. Checklists are collected and evaluated for reliability (e.g., compiled from actual specimen records versus general field guides). The comparison of checklists with the GAP maps shows: (a) the number and the identity of species predicted by GAP, and the number and identity that actually have been observed as documented by the checklist, (b) the number and identity of species predicted to occur but not observed (errors of commission), (c) the number and identity of species observed but not predicted (errors of omission), and (d) the size of the area that the checklist covers - since differences between the checklist and the GAP map may depend on the size of the area covered by the checklist. As one example, Edwards et al. (1995) compared checklists from eight national parks with the species maps from the Utah Gap Analysis Project. In this assessment, error rates tended to decrease from amphibians to reptiles to mammals to birds. The degree of agreement ranged from 60-86% for amphibians, 70-83% for reptiles, 78-85% for mammals, and 81-95% for birds. The higher amphibian error rate is attributed to the low numbers of amphibian species occurring in these parks, as well as the low number of records to work with. Omission and commission error decreased as park area increased. Commission error was generally greater than omission error, indicating that the GAP maps tended to over-predict rather than under-predict species occurrences. In addition, field surveys were conducted over two years in Idaho to check the accuracy of amphibian and reptile predicted distributions (C. Peterson, Idaho State University, Pocatello, personal communication) showing an overall accuracy of 85 percent. A sampling design for field surveys is treated by Cassidy et al. (1994) in some depth. However, while conducting field surveys to verify species distributions has its advantages, it requires several-to-many field seasons and is expensive.
In the Gap Analysis state pilot projects (Idaho and Oregon), the predicted occurrences of terrestrial vertebrate species were displayed by showing map polygons in vector format of their expected distributions. These polygons were derived by combining: (a) the recorded geographic distributions of species by county and/or region with (b) the known distribution of the habitat types of those species and (c) other variables that can further define habitat, such as temperature or elevation. In combining (a), (b), and (c) in sequence, researchers began with a very general base map showing historical records of occurrence by county or region. This base map was then modified several times with present-day habitat data to yield a cartographic probability statement of the species' current range. County-of-occurrence records were used as the base geographic unit in step (a) because most museum and agency records used them. However, counties and regions vary greatly in size, and this variation gives rise to at least two problems. First, a species' actual range may frequently extend only partly into a county while its vegetational habitat may extend throughout the county. The species' actual utilization of all vegetational habitat may be limited by competition, predation, disease, climate, or barriers, among other reasons. The result is that the species' range can be exaggerated, especially in large counties where the entire county would then be considered to be part of the species' geographic range. A second problem is the lack of consistency between states in county size, with a hundred-fold difference in size between small and large counties, even within a single state. This size variation naturally leads to inconsistencies in the resulting species distribution data and error margins. Although these do not seem to have been major limitations for the Idaho and Oregon pilot projects because of the sequential use of steps (b) and (c), the use of counties as the base geographic unit for step (a) is less than ideal. A solution to these problems is to use a constant-size "tessellator" (a means to form an equal-area mosaic) of the landscape that is smaller than the average county but not so small as to create unacceptable error rates in predicted vertebrate distributions. These equal-area geographic units can then be used to record the presence or absence of a species. White et al. (1992) developed a hexagonal tessellation of the landscape for the Environmental Protection Agency's Environmental Monitoring and Assessment Program (EPA-EMAP). The equal-area coverage they designed possesses a number of attributes that make it particularly well-suited for describing the geographic distributions of elements at map scales of 1:100,000 or smaller. After many months of consultation, the National Gap Analysis staff, working with other members of the Biodiversity Research Consortium, decided to adopt the EMAP 635 km2 hexagonal tessellation of the landscape as a cartographic unit for step (a) to show the distribution of vertebrate collection records prior to steps (b) and (c), as well as one way to show the species' predicted range after applying steps (b) and (c) (Master and Jennings 1993). Using hexagonal units to show the predicted ranges adds a new and powerful analytical tool to the conservation biologist's tool box. For example, by using this equal-area system of geographic coverage, data can be expressed as part of a global system which will allow for comparisons both within and among continents. Hexagon tessellators offer an optimum structure for implementing population and biogeographic models of species occurrence to test hypotheses about biodiversity, especially over large areas. This equal-area approach could lead to more realistic management of resources across political jurisdictions, whether county, state, or national. The hexagonal tessellation designed by White et al. (1992) can be more finely tessellated to display results from species habitat relations models at larger scales. In 1994, a pilot Biodiversity Research Consortium project in Oregon and Pennsylvania was begun. In this pilot project, state Natural Heritage Programs (NHPs) mapped hexagons-of-occurrence for vertebrate species, butterflies, skippers, freshwater mussels, trees, rare plants, and rare invertebrates. The NHPs populated the TNC Biological Conservation Database (BCD) with species distributions in the EMAP hexagonal coverage. For each species in each hexagon, the following coded information was recorded:
The pilot project was successful in proving the utility of this method. For Gap Analysis state projects that have not yet mapped vertebrate distributions, the data generated by this method will serve as a base data layer for the recorded geographic distributions of species. In states that have already undertaken Gap Analysis (e.g., Oregon), this project will both test the existing vertebrate data layer and provide additional data. For example, because of the relatively large size of some counties in Oregon, animal distributions based on the county-habitat intersections may overestimate species' within-state ranges. This can occur whenever a species' range includes part of a county rather than the entire county. In Oregon, species ranges were predicted using both methods (county and hexagon). The hypothesis that the hexagons provide a more accurate prediction range is being tested against known occurrences of species. Results will be reported out in the refereed literature. |
