One major objective of Gap Analysis is to identify areas of high biodiversity that are not managed primarily for maintaining native species and natural ecosystems (Scott et al. 1993). Identification of such areas, called "gaps", is accomplished by comparing maps of the distributions of vegetation types, terrestrial vertebrates and butterflies with maps of land stewardship and management status. This poster details the process by which maps of predicted mammalian distributions in New York are being developed for the New York Gap Analysis Project.
We obtained mammalian data in the form of museum records, game harvest records, and rabies specimen reports (see Data Sources). These data were formatted, aggregated, and edited to eliminate errors that would affect the accuracy of the final database. This process resulted in a data table containing just under 100,000 records, each describing the occurrence of a given species at a specific locality. For the purpose of comparing data from the three different sources, the aggregated data table was partitioned into three data tables: museum specimens, harvest reports, and rabies reports. A unique-records program that saved the most current site-specific record and discarded duplicate occurrence records was applied to each of the three data tables. The resulting unique records were related to an Arc/Info coverage of townships to create graphics of the distribution and diversity of unique records from each of the three data sources. These graphics and their associated data tables are compared in terms of species coverage, spatial coverage, temporal coverage, data credibility, and data volume (Part I, a., b., and c.).
The same unique-records program was applied to the aggregated data table (100,000 records). The resulting data table contained 9,004 records of the most current, unique combinations of species and township. These records were related to the township coverage to produce a graphic of the distribution and diversity of unique records within New York (Part II, a.). When this graphic is compared to the graphics in Part I, one can see how well the different sources of data compliment each other. The high species diversity of the museum data compensates for the low species diversity of the harvest and rabies data, while the excellent spatial coverage of the harvest and rabies data fill in the gaps in distribution in the museum data. While this graphic (Part II, a.) and its associated data table provide a good summary of the data for all years from all sources, the age of some records (e.g. 1850's) makes this form of the data unsuitable for representing the current distributions of mammals in New York; thus, we had to decide the maximum age of data that would be appropriate for use in our gap analysis.
Gap analysis requires the use of recently collected information to determine the geographic ranges of species within a state (Csuti and Crist 1997). The age of data that can be considered "recent" depends upon the discretion of the analyst and can be quite different from state to state. Instead of making an arbitrary decision as to which data are too old, we used a history of land-use change to aid us in the decision-making process. As patterns of land use change, the associated alterations of wildlife habitats can cause changes in the densities and distributions of mammals. Because significant land-use change can occur within relatively short time periods (e.g. decades), older records may no longer provide accurate distribution information for some species. Ideally, one would use only the most current data to develop distribution maps. The problem with this approach is that most museums have not sampled intensively in New York since the 1950's, thus there is a relative paucity of very recent museum data. Consequently, our ability to produce accurate range maps would be limited because of significant omission errors. Solving this dilemma of habitat change requires a compromise between retaining as much of the older data as possible and retaining only more recent records of species occurrence.
To arrive at a suitable compromise, we conducted a temporal analysis of land cover change and the number of most current, unique records (Part II). An important land cover change affecting wildlife populations in New York is the reversion of farmland back to woodland and forest. This type of change is important because: 1.) farms occupied the majority of New York land during the first half of this century and were largely replaced by forest during the second half of this century (Part II, Figure 1), thus large areas of the state have undergone extreme change during this century (Stanton and Bills 1996); and 2.) this change in land cover often leads to shifts in dominant wildlife species from grassland specialists in oldfields and farmlands to woodland and generalist species in shrublands and forests. Because the conversion of farmland to shrubland is one of the most dramatic changes in land cover, and because this conversion has occurred over a large percentage of the state during this century, we decided that the mammal data should be stratified by the length of time necessary for abandoned farmland to convert to shrubland. In a study of oldfield succession in New Jersey, Pickett (1982) found that woody vegetation began to attain dominance about fifteen years after the farm under study was abandoned. Using this information as a guide, we stratified the data related to the "All Years" graphic (Part II, a.) by fifteen-year intervals to produce seven new graphics (Part II, b.-h.) and associated data tables.
Because farmland in New York declined by only 13% between 1965 and the 1990's (Part II, Figure 1), we believe that data going back as far as 1965 could be considered "recent" for our purposes. The "1965 to Present" data set (Part II, g.) would allow retention of almost 76% of the total unique records (Part II, Figure 2). After comparing distribution maps for several individual species using this data set, we discovered that it was inadequate because it contains few records for entire regions of the state, such as Long Island, and use of it would lead to gross errors in distribution maps. Hence, we decided to go back further in time to evaluate the "1950 to Present" data set (Part II, f.). This data set has two distinct advantages over the "1965 to Present" data set: 1.) the distribution maps for the same individual species do not contain substantial omission errors; and 2.) it allows retention of over 86% of the total unique records (Part II, Figure 2), a 10% gain over the previous data set. However, a main disadvantage is that farmland declined by about 25% between 1950 and the 1990's (Part II, Figure 1), thus there is an additional 12% change in land cover to consider. Notwithstanding the effects that the additional change in land cover might have had on mammal distributions, we believe that the "1950 to Present" data set will produce more useful distribution maps than the "1965 to Present" data set. Because the "1950 to Present" data set appeared adequate for gap purposes, there was no need to go back even further in time to evaluate other data sets. The "1950 to Present" data set was judged to be the best compromise between maximizing data retention and maintaining data reliability; thus, it was chosen for use in developing range maps for all native mammals in New York.
Part III shows range maps for selected species based on reported occurrences contained in the "1950 to Present" data set (Part II, f.). These maps provide insight about where species are known to occur within New York. These known distributions were used in conjunction with habitat relation models, which link species to habitat characteristics such as vegetation alliance and elevation, to produce maps of predicted distribution for the same selected species (Part IV). In the next step, the predicted distribution maps will be subjected to review by mammal experts, and appropriate edits will be made to create the final distribution maps. The final distribution maps will be compared with maps of the distributions of vegetation types, other terrestrial vertebrates, and butterflies for assessment of overall biodiversity in New York.