Identifying habitat for species with special ecological requirements can be a challenging task when procedures are based on remotely sensed data. I used georeferenced locations of red-cockaded woodpecker (Picoides borealis) cavity trees to evaluate the effectiveness of Landsat Thematic Mapper (tm) data and a digital elevation model in identifying oldgrowth pine forests that provide nesting habitat for this species. Remotely sensed data associated with active cavity trees (N=142) and polygons surrounding active cavity clusters (N=179) were compared to locations with unsuitable habitat (N=1000). Elevation was the best predictor of woodpecker locations, but some TM bands improved classifications slightly. The best classification (overall accuracy=74%, kappa=0.45) was based on an elevation mask and transformed TM data associated with the red, blue, and green TM bands. TM bands were transformed using linear stretching procedures and neighborhood statistics computed using a 5-by-5 pixel window. Accuracy might be further improved by analyzing patch size and shape characteristics, but such analyses would be complex and likely still fall short of common accuracy standards. Other procedures to identify potential habitat, such as aerial photo interpretation, are more appropriate for this species, and any attempt at habitat identification must allocate sufficient time for extensive field surveys.
