Effects of species ecology and urbanization on accuracy of a cover-type model: A test using GAP analysis

Abstract

Models of vertebrate distributions based on dominant vegetation cover or land-use classification are commonly used for conservation planning, but these models may be inappropriate for species that choose sites based on criteria other than land cover or within urban areas that are not adequately described by cover-type alone. We compared the accuracy of predicted occupancy of birds for a set of cover-type models—Alabama Gap Analysis Program's (ALGAP) vertebrate distribution maps—between an urban and a rural landscape in east-central Alabama. We performed analysis at two scales of investigation—0.03-km2 point-count surveys or 28.26-km2 landscapes—using point counts conducted during summers 2004–2006. We tested ALGAP's ability to predict the occupancy of habitat by birds grouped by life-history parameters: migrant, resident, insectivore, carnivore, and omnivore, forest dweller, and cavity nester. ALGAP performed well at the scale of entire landscapes but poorly at the scale of individual point counts. At the point-count scale, ALGAP was most accurate for species requiring interior forest conditions. At the landscape scale, ALGAP was more accurate in the rural landscape than the urban landscape, and it had higher commission errors in the urban landscape. Variation in the ability of ALGAP to predict species occupancy was likely due to (1) poor model performance when applied to species that choose sites using criteria other than cover type and (2) the inadequacy of ALGAP to describe a heterogeneous urbanized landscape. Our results highlight pitfalls of using land cover information to model species distributions in situations where it may be inappropriate.