Spatial modeling of rangeland potential vegetation environments

Potential vegetation environments (e.g., habitat types, range sites, ecolog ical sites) are important to land managers because they provide a conceptua l basis for the description of resource potentials and ecological integrity . Efficient use of potential vegetation classifications in regional or subr egional scale assessments of ecosystem health has been limited to date, how ever, because traditional ecological unit mapping procedures often treat su ch classifications as ancillary information in the map unit description. Ac cordingly, it is difficult, if not impossible, to describe the precise loca tion, patch size, and spatial arrangement of potential vegetation environme nts from most traditional ecological unit maps. Recent advances in remote s ensing, geographic information systems (GIS), terrain modeling, and climate interpolation facilitate the direct mapping of potential vegetation throug h a predictive process based on gradient analysis and ecological niche theo ry. In this paper, we describe how a predictive vegetation mapping process was used to develop a 30 m raster-based map of 4 grassland, 5 shrubland, an d 6 woodland habitat types across the Little Missouri National Grasslands, North Dakota. Discriminant analysis was used in developing this potential v egetation map based on 6 primary geographic information system themes. Geoc limatic subsections and remotely sensed vegetation lifeform maps were used in predictive model stratification. Terrain indices, LANDSAT satellite imag ery, and interpolated climate information were used as independent (predict or) variables in model construction. A total of 616 field plots with known habitat type membership were used as dependent variables and assessed by a jackknife discriminant analysis procedure. Accuracy values of our map range d from 54 to 77% in grasslands, 62 to 100% in shrublands, and 70 to 100% in woodlands dependent on geoclimatic subsection setting. Techniques are also described for generalizing the 30 in pixel resolution habitat type map to appropriate ecological unit maps (e.g., landtype associations) for use in e cosystem health assessments and land use planning.