ESD logo


In order to determine potential conflicts between training or other land uses and natural resources, it is necessary to know where those natural resources are located. Historically, mapping of resource locations has involved extensive field work over large areas even though the resource may be located patchily within the areas. Geographic information system (GIS) technology provides the ability to construct models of habitat that rely on existing or readily obtained information (e.g., remotely sensed images, soil surveys, digital elevation models, geological surveys, topographic maps, etc.). Such models offer the possibility of being able to minimize field work and to focus field activities in much smaller areas, and GIS-based models are easily updated as new information becomes available.

We used a combination of inductive and deductive GIS modeling approaches to predict potential habitat for two bird species, Henslow's sparrow (Ammodramus henslowii) and cerulean warbler (Dendroica cerulea), and for limestone cedar barren communities at Fort Knox (Kentucky). Both birds were previously listed as Federal category 2 candidate species, and Henslow's sparrow is listed as a special concern in Kentucky. The cedar barrens habitat at Fort Knox contains several state-listed and previous Federal candidate species of plants. In this hybrid approach we make use of a number of spatial data layers, including soils and geology, to produce robust models that do not rely so heavily on the accuracy of any given layer.

Stoms et al. (1992) distinguish between inductive habitat modeling approaches, in which characteristics of locations where species occur are generalized to the rest of the management area, from deductive rule-building approaches, in which inferences are made from the general to a particular case. Deductive approaches are determined a priori, and try to anticipate the organisms by explicitly choosing habitat criteria which are believed to be important. In inductive approaches, habitat choices of a subset of the organisms are observed, and the chosen habitat characteristics are extrapolated to wider areas.

GIS-based habitat models are usually based on an exclusively deductive or inductive approach, but few habitat modeling studies have integrated both techniques. Many habitat models are based on deductive rules involving land cover type and other characteristics. For example, Clark et al. (1993) developed a deductive multivariate model of female black bear (Ursus americanus) habitat in the Ozark National Forest based on forest cover and several topographic and spatial parameters. Rudis and Tansey (1995) modeled black bear habitat on a regional basis for the entire southeastern United States using deductive rules based on Forest Inventory Analysis surveys from the U.S. Forest Service.

Inductive methods generalize from sites known to be habitat to the rest of the map. Census data form the foundation for inductive habitat prediction. Homer et al. (1993) used Landsat Thematic Mapper (TM) data to model sage grouse (Centrocercus urophasianus) habitat in Northern Utah. Coker and Capen (1995) used aerial surveys of the Green Mountain National Forest along with field census data to develop an inductive model of cowbird (Molothrus ater) use of disturbance patches in the forest. Knick and Dyer (1997) analyzed remotely-sensed shrub vegetation within 1 km of black-tailed jackrabbit (Lepus californicus) sightings to inductively predict habitat and estimate loss to large-scale fires.

Lauver and Whistler (1993) used a hierarchical inductive classification of Landsat TM data to identify native grasslands in eastern Kansas, USA. Discriminant analysis of ground occurrence data was extrapolated to distinguish high-quality from low-quality grasslands. Seventy-seven previously unknown natural grassland areas were identified, nine of which contained populations of the Federally-threatened Mead's milkweed (Asclepias meadii).

In one of the few habitat modeling studies to have integrated both deductive and inductive approaches, Sperduto and Congalton (1996) used GIS to predict potential habitat for the small whorled pogonia (Isotria medeoloides), the rarest orchid in eastern North America north of Florida. Using a weighted hybrid model which included slope, aspect, and soil characteristics with Landsat TM Band 4 reflectance signatures generated from known orchid locations, they correctly predicted 78% of the known occurrence sites for this orchid in New Hampshire and Maine.

Lowell and Astroth (1989) also used a hybrid GIS approach to predict cedar barren communities within the Hercules Glades Wilderness Area, Missouri at five intervals over a 48-year period. They identified physiographic factors correlated with the presence of glades using a modified Chi-squared analysis, which indicated a strong positive association with shallow Gasconade soil, elevations from 305-365 m, and southern and southwesterly aspects. This model of physiographic and edaphic factors allowed them to predict 93% of the glades known to be present in their study area.


Clark, J. D., J. E. Dunn, and K. G. Smith. 1993. A multivariate model of female black bear habitat use for a geographic information system. J. Wildl. Manage. 57(3):519-526.

Coker, D. B., and D. E. Capen. 1995. Landscape-level habitat use by brown-headed cowbirds in Vermont. J. Wildl. Manage. 59(4):631-637.

Homer, C. G., T. C. Edwards, R. D. Ramsey, and K. P. Price. 1993. Use of remote sensing methods in modelling sage grouse winter habitat. J. Wildl. Manage. 57(1):78-84.

Knick, S. T., and D. L. Dyer. 1997. Distribution of black-tailed jackrabbit habitat determined by GIS in southwestern Idaho. J. Wildl. Manage. 61:(1):75-85.

Lauver, C. L., and J. L. Whistler. 1993. A hierarchical classification of Landsat TM imagery to identify natural grassland areas and rare species habitat. Photogrammetric Engineering and Remote Sensing 59(5):627-634.

Lowell, K. E., and J. H. Astroth, Jr. 1989. Vegetative succession and controlled fire in a glades ecosystem: a geographical information system approach. Int. J. Geographical Information Systems 3(1):69-81.

Rudis, V. A., and J. B. Tansey. 1995. Regional assessment of remote forests and black bear habitat from forest resource surveys. J. Wildl. Manage. 59(1):170-180.

Sperduto, M. B, and R. G. Congalton. 1996. Predicting rare orchid (small whorled pogonia) habitat using GIS. Photogrammetric Engineering and Remote Sensing 62(11):1269-1279.

Stoms, D. M., F. W. Davis, and C. B. Cogan. 1992. Sensitivity of wildlife habitat models to uncertainties in GIS data. Photogrammetric Engineering and Remote Sensing 58:843- 850.

Return to top of this page

Return to the Model Categories page

Return to the Web Site Map

Return to Ecological Modeling home page