Kelly L. Wetteroff, Dr. Ronald D. Drobney, Timothy L. Haithcoat

EVALUATING VERTEBRATE DISTRIBUTION MODELING
WITHIN A GIS FRAMEWORK

Abstract

We will present a methodology for creating predicted vertebrate ranges for Gap Analysis utilizing species specific perceptions and spatial constraints based on measures of the environment and landscape within which they live. This methodology utilizes many more capabilities of GIS as well as incorporating statistical measures of landscape to refine the spatial extent of a species range. We will include discussions of scale, patch, matrix measures, and landscape context and structure, as they relate to the selection and modeling of vertebrates. Our pilot study covers 4 counties in mid-Missouri. We have collected statewide information on distributions and habitat preferences for many species within each terrestrial taxon for the development of a comprehensive statewide digital database. This database will be used in association with the Missouri land cover map to determine areas of suitable habitat within the known geographical range of each species modeled.

Introduction

We have been developing and evaluating new methodologies for modeling vertebrate distribution as part of the Missouri Gap Analysis Program. By way of a brief introduction, Gap Analysis is a federal program, implemented at the state level, and intended to be a rapid and coarse overview of current biodiversity protection status (Scott et al. 1993). Gap Analysis assesses the distribution of wildlife species by analyzing existing species distributions and the protection status afforded to them from the land ownership or management practices on associated lands. Gap Analysis originated from a lack of success with the species-by-species approach to conservation, which ignores the primary reason for the loss of biodiversity, the continual loss of habitat and the fragmentation of natural landscapes (McNeely et al. 1990, Bridgewater 1993).

The Gap Analysis Program incorporates three data layers to reach an assessment of protected and unprotected areas. These layers are a) actual vegetation - derived from satellite imagery and ancillary data,
b) wildlife species habitat and range information, and c) land ownership/management status. The GAP theory in combining these layers is that a species has a high probability of being present in areas of suitable habitat within the known distributional limits of the species (Csuti 1993).

Several specific pieces of information are involved in creating vertebrate distribution models. These are:

1. Habitat association data; correlating habitat types identified within the literature with the vegetation types discriminated on the vegetation image map.
2. Wildlife - habitat matrix; recording and modeling which components of the habitats mapped are associated with which vertebrate species.
3. Known range information; maps of the known distributional limits of the vertebrate species and any associated locational or breeding information.

The development of vertebrate models requires the compilation of a database of wildlife species and their preferred habitats. Typical sources for this information include local expert information and records published in the literature. We were fortunate to have access to the Missouri Fish and Wildlife Information System (MoFWIS), a wildlife database managed and maintained by the Missouri Department of Conservation (MDC). We have relied on this database for information such as species-habitat associations and average home range size. When this information was modeled against the map of habitat types, we were able to designate areas of potentially suitable habitat for each species under consideration.

Producing maps of predicted vertebrate distributions also requires a comprehensive database of known vertebrate ranges. Vertebrate ranges are usually generalized from records initially recorded as point locations. Typical sources of point location data are: museum records, published research and other field observations, Natural Heritage Program databases, and Breeding Bird Atlas databases. These point sources must be generalized into a polygon (or polygons) representing the range of the species. In Missouri, for both herptile and mammalian distributions, we relied on published generalized range maps from regional experts. We digitized these maps into our database. Avian range information was obtained through a cooperative effort with MDC. These records were already generalized into distribution by USGS quadrangle.

Methodology

The basic Gap Analysis vertebrate model is quite simple: if a species uses a particular habitat type, all polygons of that type are included in a map of suitable habitat. Other digital layers may be used where available and appropriate, such as elevation, climate, and soil, to more realistically represent areas of suitable habitat. The second step in Gap Analysis vertebrate modeling is a simple overlay of the habitat map (the vegetation map) with a map of each species' distributional limits to result in the elimination of habitat polygons outside the known range of the species. A final species map is an illustration of all suitable habitat areas within the known geographic range of the species. However, we feel that there are many additional measurable habitat components which can be included to provide a more accurate species map.

In the currently developing field of landscape ecology, attempts are being made to identify important landscape components affecting wildlife habitat and to quantify their structure. Many of these habitat elements and measures may be easily included in the models through the use of GIS. Some of these significant habitat variables are: juxtaposition and interspersion of habitat types, identification of core areas, contrast between edge and adjacent area, the proportion of different habitat classes within a specified area, distance measures (i.e., from roads/human habitation or from water), and area measures.

In order to perform this species modeling, we first needed a base habitat map. This was developed from a vegetation map derived from classified Landsat TM imagery. Within ArcInfo, we decided to use the GRID module for our vertebrate modeling. We chose to use the grid environment rather than the polygon environment because grids use a smaller amount of storage space and process much faster than polygon coverages.

From the classified satellite land cover map, each different habitat type was converted into a separate grid. Examples of the habitat grids are:

Forest - solid/interior, linear, and mixed with cedar
Grass - tall and short grass, where differentiated
Agriculture
Urban
Water bodies
Stream corridors

We also incorporated coverages of the roads, rivers, and streams in the study area. Each of these original arc files was converted into a grid for use with the models as well. We also identified certain elements of each vegetation class which might be important for wildlife habitat. For example, many species are affected (both positively and negatively) by habitat edges. In order to identify a forest-grassland edge, we started with the upland forest grid and the upland grassland grid. All habitat grids were reclassified into binary grids from the original vegetation grid. In each of these grids, cells of the desired habitat type were identified with a value of 1; all other cells were identified as 0. Beginning with these single habitat type grids, we then used the EXPAND function to extend the boundary of each habitat type. We added these two expanded grids together; adding the upland forest expanded grid and the upland grassland expanded grid resulted in cells with three values (0,1, and 2). We then reclassified the resulting grid so that cells with the value of 2 became a value of 1 and all other cells were reclassified to 0. This grid now represented an upland forest-grassland ecotone of 30 meters. Similar functions can be used to identify, for example, interior habitat areas, and thus mitigate any edge effect, and also to identify habitat areas of a certain size, of concern for species with minimum area requirements. Each species model involves a simple additive process and can be developed by including any pertinent habitat parameters. To allow the development of these models, all of the possible habitat variables have been converted into binary grids, with the desired habitat element in each grid having a 1 value and all other grid cells having a 0 value.

We have also included an assessment of the landscape from an individual species' perspective. In other words, we have attempted to view the landscape as one of the modeled vertebrate species might. We used MoFWIS to identify average home range sizes for as many species as possible. Based on the wide variety of home ranges, we identified 5 major levels of home range size: 120, 420, 840, 2010, and 5010 m. (This is the value for one dimension of the home range. The actual area of the home range is this value, squared.) We generated four grids for each of these home range sizes. Each of the four generated grids was offset one-half the distance of one side of the home range. For example, the origin of grid 120a was at (0,0); grid 120b, (-60, 0); grid 120c, (-60, -60); and grid 120d, (0,-60). These empty grids were generated in Arc, and then each was converted into a grid, with a resolution of 30 m.

We used these overlapping home range units with the additive species model described above. After developing the model for a species, which involved adding several different habitat grids, we then performed a ZONALMEAN in Grid on the resultant model, for each of the four home range grids (a, b, c, and d). These four grids were then added together for a cumulative grid; this grid was then divided by 4, to obtain a new grid with an average value for each cell. These values were typically very small, and thus, before converting the floating point grid into an integer grid, we multiplied the average grid by 100, in order to retain the significance of the values. This final grid for each species has a variety of values representing habitat of differing suitability for the species under consideration. Higher cell values represent areas of higher suitability relative to areas with lower cell values. By using this modeling approach, we have been able to produce refined maps of potentially suitable habitat with rankings illustrating relative measures of suitability.

We intend to perform the habitat modeling with both the standard Gap Analysis procedure, and with the additional GIS operations described above. We will then assess results from the two types of models to determine if the additional grid analyses provide significant information about predicted wildlife distribution and enhance the standard Gap Analysis.

References

Bridgewater, P. B. 1993. Landscape ecology, geographic information systems and nature conservation. Pages 23-36 in R. Haines-Young, D. R. Green, and S. Cousins, eds. Landscape Ecology and Geographic Information Systems. Taylor and Francis. Philadelphia. 288 pp.

Csuti, B. 1993. Methods for developing terrestrial vertebrate distribution maps for gap analysis (Data Layers). Pages 2.1-2.52 in A Handbook for Gap Analysis. National Biological Survey Gap Analysis Program, Version 1. USFWS, Idaho Cooperative Fish and Wildlife Research Unit, University of ID.

McNeely, J. A., K. R. Miller, W. V. Reid, R. A. Mittermeier, and T. B. Werner. 1990. Conserving the World's Biological Diversity. International Union for Conservation of Nature and Natural Resources, World Resources Institute, Conservation International, World Wildlife Fund - US and The World Bank. Gland, Switzerland. 193 pp.

Scott, J. M., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves, H. Anderson, S. Caicco, F. D'Erchia, T. C. Edwards, Jr., J. Ullman, and R. G. Wright. 1993. Wildl. Mon. 57(123):1-41.

Author Information

Kelly L. Wetteroff
Graduate Research Assistant
School of Natural Resources - Division of Fisheries & Wildlife
University of Missouri
20 Stewart Hall
Columbia Missouri, 65211
phone: 573-882-1404
fax: 573-884-4239
c653107@showme.missouri.edu

Timothy L. Haithcoat
Senior Research Specialist
Department of Geography / Geographic Resources Center
University of Missouri
16 Stewart Hall
Columbia Missouri, 65211
phone: 573-882-1404
fax: 573-884-4239
grctlh@showme.missouri.edu

Ronald D. Drobney
Cooperative Fish and Wildlife Research Unit
University of Missouri
112 Stephens Hall
Columbia Missouri, 65211
phone: 573-882-3436