USING LANDSCAPE-LEVEL DATA TO PREDICT THE DISTRIBUTION OF BIRDS ON A MANAGED FOREST: EFFECTS OF SCALE

Abstract

Selection of scale is critical when investigating ecological processes on landscapes because different patterns emerge in spatial data at different scales. Landscape studies commonly identify a single scale, or spatial extent of data, for assessing broad-scale habitat characteristics, without regard for the sensitivity of spatial data to the scale at which they are measured. An incorrect selection of scale can lead to misleading or erroneous inferences about how animals are associated with coarse-grained habitat characteristics. We developed and compared three statistical models for predicting presence of selected bird species inhabiting a managed forest in South Carolina: a model based only on microhabitat characteristics, a model based only on landscape characteristics (summary statistics of forest age and type calculated at different spatial scales) derived from GIS data, and a model that combined microhabitat and landscape characteristics. In general, landscape models (Somer's D = 0.61 ± 0.16; mean ± 1 sd) worked as well as microhabitat models (D = 0.61 ± 0.14), and combining the two types provided only a slight improvement in the explanatory ability of the models (D = 0.62 ± 0.18). Models for Neotropical and short-distance migrants had the highest fit to field data, whereas models for resident species had relatively poor fit. We refined our landscape models according to known or hypothesized information from the literature to improve their generality, and we tested their ability to predict presence of the same species on a second, independent data set collected on a different managed forest nearby in South Carolina. In general, landscape models were able to predict the distribution of selected birds on the second forest well (D = 0.46 ± 0.32), although overall model fit was somewhat lower than for the first forest (D = 0.61 ± 0.16). Model fit was greatest for Neotropical and short-distance migrants, and poorest for residents. Model fit did not vary according to successional status, but did vary with habitat specialization; model fit was highest for habitat specialists and lowest for generalists. Our results suggest that, in general, coarse landscape characteristics are most important to migratory bird species that are limited in the number of habitats they can use for breeding. For species with adequate fit of landscape models, we assessed relationships between landscape scales associated with habitat variables within each model and ecological characteristics. Scale did not vary with migratory status, successional status, or habitat specialization and appears to be a function of the unique natural history of a species. Scale was correlated with hypothesized area sensitivities for some forest interior species, but not all; some early-successional species also appeared to be area sensitive. We conclude that no single scale is appropriate for assessing landscape associations across all bird species, or across general ecological guilds of species. Our modeling approach provides forest managers with a robust, biologically based approach to assessing the effects of forest management on birds across an entire landscape, using only GIS data.