A test of the Swainson's warbler habitat suitability index model

We tested a recent Habitat Suitability Index (HSI) model proposed for the Swainson's warbler (Limnothlypus swainsonii; a species of conservation concern in the southeastern United States) using data on habitat, bird occupancy, and abundance collected at White River National Wildlife Refuge, Saint Francis National Forest, and Big Island in Arkansas between 2005 and 2010. The Swainson's warbler HSI includes a combination of 6 variables: landform, landcover, successional age class, forest patch size, proportion forest in a 1-km radius, and small-stem density (<2.5 cm diam breast ht; these variables are combined to produce scores ranging 0-1, representing unsuitable to optimal habitat. Mean HSI scores were significantly lower (x=0.65, n-99) for sites occupied by Swainson's warblers compared with unoccupied locations (x=0.78, n=177) at White River National Wildlife Refuge and Big Island, contrary to our expectation for a useful model. We also found that Swainson's warbler abundances were not correlated with SI scores both at Saint Francis National Forest and at White River National Wildlife Refuge. Our tests of the current Swainson's warbler HSI model demonstrate that it was not an effective predictor of Swainson's warbler occupancy and was not a reliable indicator of warbler habitat at the scale of our study areas. We propose a new suitability function called proportion elevation that takes relative elevation into account at potential sample locations that may improve the model. Specifically, this function should index the likelihood a site is infrequently flooded and should support the development of a thick understory, a habitat feature that has been demonstrated to be important to Swainson's warblers.