Inferring habitat quality and habitat selection using static site occupancy models

Abstract

When evaluating the ecological value of land use within a landscape, investigators typically rely on measures of habitat selection and habitat quality. Traditional measures of habitat selection and habitat quality require data from resource intensive study designs (e.g., telemetry or mark-recapture). Often, managers must evaluate ecological value despite only having data from less resource intensive study designs. In this paper, we use occupancy data to measure habitat quality and habitat selection response for the Puerto Rican Vireo, an endemic songbird whose population growth is depressed by brood parasitism from the Shiny Cowbird. We were interested in how vireo habitat quality and vireo habitat selection varied among three land uses (forest, shaded coffee plantations, and sun coffee) in Puerto Rico. We estimated vireo occupancy probability as a measure of habitat selection, and the probability of cowbird occurrence given vireo presence as a measure of habitat quality. To estimate the latter, we explored different ways of modeling the occurrence of the two species jointly, and compared these models to independent models using measures of predictive performance. Vireos preferentially selected forested sites and shaded coffee sites over sun coffee sites. By our measure of habitat quality, either type of coffee plantation was poor quality, and the forested sites were high quality. This suggests that shade coffee may be an ecological trap for the vireo in our study area. One joint model performed best by our measures of predictive ability, thus showing that the cowbird may not occur independently of the vireo. Vireo population dynamics in our study area may benefit from having large amounts of forest relative to coffee plantations. Incorporating species interactions into occupancy models has the potential to improve monitoring for conservation.