High-resolution image texture as a predictor of bird species richness

We tested image texture as a predictor of bird species richness in a semi-arid landscape of New Mexico. Bird species richness was summarized from 10-min point counts conducted at 12 points within 42 plots (108 ha each) from 1996 to 1998. We calculated 14 first- and second-order texture measures in eight different window sizes on a set of digital orthophotos acquired in 1996. For each of the 42 plots, we summarized mean and standard deviation of each texture value within multiple window sizes. The relationship between image texture and average bird species richness was assessed using linear regression models. Single image texture measures such as the standard deviation described up to 57% of the variability in species richness. Coupling multiple measures of texture or coupling elevation with a single texture measure described up to 63% of the variability in bird species richness. Models incorporating two measures of texture and coarse habitat type described 76% of the variability in bird species richness. These results show that image texture analysis is a very promising tool for characterizing habitat structure and predicting patterns of species richness in semi-arid ecosystems. This method has several advantages over methods that rely on classified imagery, including cost-effectiveness, incorporation of within-habitat vegetation variability, and elimination of errors associated with boundary delineation.