PARENT SESSION
Contributed Oral Session 64: GIS / Remote Sensing and Landscape Ecology
Tuesday, August 9, 1:30 PM - 5:00 PM, Meeting Room 519 B, Level 5, Palais des congrès de Montréal
Using image texture analysis to predict breeding birds' response to fine scale habitat variability in semi-arid environments.
St-Louis, Véronique*,1, Pidgeon, Anna1, Radeloff, Volker1, Hawbaker, Todd1, 1 University of Wisconsin-Madison, Madison, WI
ABSTRACT- Land cover classifications of remotely sensed imagery are widely used to study birds' response to habitat. However, this approach has inherent limitations in semi-arid ecosystems, where patch boundaries are often "soft". Our goal was to use unclassified images to: a) predict breeding birds' response to fine scale structural variability in desert shrub- and grass-dominated habitats, and b) assess individual species response to the grain and the spatial scale at which structural variability is analyzed. We hypothesized a stronger response to grain and scale for specialist species when compared to generalists. Breeding bird data were collected at 42 plots (108 ha each) in the McGregor Range of Ft. Bliss, NM during three breeding seasons from 1996 to 1998 via point counts. Texture analysis was performed on orthophotos and Landsat TM imagery, with spatial resolution (grain) of <1-m and 30-m respectively. We calculated several texture measures (variation in grey tones, contrast and range) by using moving windows of different sizes (scale of analysis) ranging from 3x3 to 101x101 pixels. Regression models were used to relate bird species richness and abundance to texture. Relatively simple texture measures such as standard deviation in grey tones were found to be good predictors of species richness and abundance. As hypothesized, specialist species (e.g., Pyrrhuloxia, Cardinalis sinuatus) response to texture measures depended strongly on analysis window size as opposed to generalist species (e.g., Black-throated Sparrow, Amphispiza bilineata). These results emphasize the importance of the grain and scale of analysis in evaluating species' response to their habitat, especially for specialist species. Results also suggest that texture analysis of remotely sensed data is a viable alternative to the common approach based on land cover classifications when studying breeding birds' response to fine scale habitat variability in semi-arid environments.
Key words: remote sensing, image texture analysis, migratory birds, semi-arid environments