A map of the Idaho Current Land Cover layer can be viewed or downloaded from the ID-GAP Maps page.

For ID-GAP, land cover was mapped in two sections. Redmond et al. (1996) at the Wildlife Spatial Analysis Lab (WSAL), University of Montana, mapped the northern part of the state. Homer (1998), at the Utah State University Remote Sensing/GIS Laboratories, mapped the southern part of the state as part of the Wyoming mapping initiative. Contracting with two different remote sensing labs, which were already mapping vegetation in adjacent states, expedited the development of Idaho's vegetation layer for Gap analysis. It also created a minimally disjunctive land base on which to conduct subsequent research. Although the mapping endeavors were conducted independently, Homer (1998) designed a vegetation classification system that would compliment the earlier work of Redmond et al. (1996).

The Northern Idaho vegetation map was created from Landsat TM scenes and stored in raster data sets comprised of a series of 7 ARC/INFO grids (one per TM scene covering Northern Idaho). The database was built through a two-stage classification involving both unsupervised and supervised procedures. First, for each TM scene, an unsupervised classification of pixels was conducted. This pixel classification, based on Euclidean distance calculations, was designed to maintain patterns observed in a color composite of channels 4, 5, and 3. The resultant spectral classes were then regrouped and merged to 2 ha MMU (>22 pixels). Next, a raster database was constructed in ARC/INFO: base regions (or raster polygons) were delineated, and attributes for each region were collected. Meanwhile, 7.5 minute quadrangles were selected for field sampling in 1994-95 by the U.S. Forest Service, Northern Region. These ground-truth plots were combined with plots from pre-existing sources and passed to the WSAL, where they underwent a series of logical and positional tests to verify their accuracy and utility for supervised classification purposes. In all, 17,854 plots were compiled in the ground-truth database. 80% of these plots were used in the subsequent supervised classification, and 20% were used to conduct the accuracy analysis for the classification system. The supervised classification system assigned cover type labels using the 'Nearest Member of Group' classifier. Decision rules were applied where necessary in assigning labels to vegetation, size class, and canopy cover. The riparian vegetation was mapped through a separate process. Using digital elevation data, predicted riparian zones were delineated, then spectral classes were selected to represent riparian vegetation within the zones at 30m pixel resolution.

The effective resolution of vegetation cover-types identified with Landsat TM mapping is largely based on two criteria; (1) the kind of environment the mapping is occurring in, and (2) the file pixel sample size. For southern Idaho mapping zones are efforts to optimize these criteria and gain desired resolution within acceptable budgetary and time lines. A mapping zone is defined as an independent mapping project area. (Vegetation training sites and classification are applicable only to this area). With mapping zones, an effort is made to contain spatially similar ecological areas within a reasonable sample of TM pixels. It was determined that nine mapping zones would optimize this mapping effort. In each zone a master scene was chosen, and surrounding scenes slaved into the master scene. A two-step approach of atmospheric standardization and histogram adjustment was used to mosaic the TM imagery. Cover-type class definition was based first on correlation with previous Utah and Nevada classifications, and second, with the classification scheme generated by the University of Montana. Signatures in each mapping zone were classified using the ERDAS (TM) ISODATA algorithm (Tou and Gonzalez 1974) to generate unsupervised spectral clusters. An iterative review of the clustering process was used to identify the optimum number of spectral clusters needed to characterize land-cover variation in each mapping zone Cover-type modeling followed protocol developed by Homer et. al. (1997) and consisted of two phases: (1) statistical association of spectral classes with cover-types, and (2) ecological modeling based on ancillary information. 

The resulting land cover data set consists of 82 classes and is the highest resolution, continuous land cover map yet to be produced for Idaho. Idaho's most extensive vegetation community is comprised of Basin Big Sagebrush (Artemisia tridentata) and Wyoming Big Sagebrush (Artemisia tridentata wyomingensis) across southern Idaho. It covers 34,787 square kilometers or 16.08% of Idaho's land. All sagebrush and shrub-steppe types combined constitute 33% of the Idaho landscape. Agriculture ranks second in land area with 29,029 square kilometers or 13% of land cover. Grassland and meadow vegetation communities occupy 11% of the Idaho landscape, with Perennial Grasslands comprising 46% of that area. Douglas-fir is the most common forest type (7%) in Idaho, and no other single forest species or forest community occupies more than 5% of the state landscape. The total forest area is 37% of the Idaho landscape. Riparian, wetlands, and marshes cover 2% of Idaho's landscape and are categorized in seven classification codes. Shrub dominated riparian occupies the largest area with .87% of the total mapped riparian/wetland distribution. The combined sand and rock classifications occupy 2% of the landscape with the greatest portion of that distribution seen in exposed rock.

Assessed accuracy measures of the land cover map vary greatly between areas. Particular attention should be paid to the sample size for each cover type when interpreting the results. For the 5 scenes combined to create the north Idaho land cover map, producer’s accuracy for those comparisons acceptable or better (3 or greater in the fuzzy matrix) ranged from 53.35% to 71.23%. Total percent correct measures for southern Idaho mapping units ranged from 65.5% to 79.3%. Overall percent correct for the southern Idaho land cover classification was 69.3%. Overall total percent absolutely correct for the Idaho Land Cover Classification was 50.15%. Estimated kappa value for the Idaho Land Cover Classification was 0.4942.