Department of Defense natural resource managers need information on where specific bird species of conservation concern are most likely to occur, and where hotspots of migratory species richness are located, in order to successfully implement installation-specific Integrated Natural Resource Management Plans.  The objective of this project was to develop an approach to characterizing the patterns of presence and abundance of bird species in open-canopy ecosystems.

The approach integrates landscape-level analysis with spatially detailed habitat information, using remotely sensed image texture, which is derived from unclassified imagery. Image texture was calculated from digital aerial photos, from Landsat Thematic Mapper satellite imagery, which provides data on individual bands of the electromagnetic spectrum, and from a derived product, the Normalized Difference Vegetation Index.

In the Chihuahuan Desert ecosystem of Fort Bliss, New Mexico, Normalized Difference Vegetation Index alone accounted for up to 82% of the variability in species richness, while image texture derived from aerial photos was the strongest single predictor of species richness (capturing 52% of variation) in the grassland savanna-woodland ecosystem of Fort McCoy, Wisconsin. Maps of avian abundance and occurrence were developed from models using image texture at both installations. The mapping effort revealed that texture measures account for variability within landcover classes, reflecting finescale patterns of species abundance and occurrence that are not apparent in maps based on models using landcover class data only, while retaining a broad-extent perspective. In the Fort McCoy ecosystem, the near-infrared band from Landsat Thematic Mapper imagery accounted for 74% of ground-measured vertical vegetation structure, which is strongly associated with bird distribution patterns. In a focal study on Fort Bliss of the loggerhead shrike, a species of conservation concern, models based on local (i.e., ground measured) and intermediate-scale (i.e., image texture) but not broad-scale (i.e., land cover class) habitat variables best explained shrike occurrence. Finally, the research group conducted an analysis of phenological variation of texture measures in three North American biomes and found that interseasonal variability is strong, which both offers opportunities and presents challenges. The degree to which texture measures were robust to seasonality and scale of analysis was quantified.

Image texture is a useful measure in models of habitat. Models based on image texture performed equal to or better than models based on classified habitat maps for characterizing habitat use by birds, across broad extents. This project has highlighted the potential to integrate remotely sensed measures of habitat structure in habitat models.