A method for detecting undervalued resources with application to breeding birds.

Anthropogenic changes to ecosystems can decouple habitat selection and quality, a phenomenon well illustrated by ecological traps in which individuals mistakenly prefer low-quality habitats. Less recognized is the possibility that individuals might fail to select high-quality habitat because of the absence of some appropriate cue. This incorrect assessment of resource quality can lead to relatively high-quality resources being undervalued, whereby they support fewer individuals than optimal. We developed a habitat selection model to predict the expected patterns in patch-level density, fitness, and individual quality derived from either accurate assessment of habitat quality or from undervaluing of habitat patches (i.e., quality is not correctly assessed). Unlike previous habitat selection models, we explicitly and simultaneously incorporated variation in both individual and habitat quality into our estimates of realized fitness. Although multiple mechanisms can reduce patch-average density, fitness, and individual quality in less preferred patches, only undervaluation results in the occupation of higher-quality territories by similar-quality individuals in less preferred vs. preferred patches. We then looked for evidence of undervaluation in our seven-year data set of Acadian Flycatchers (Empidonax virescens) occupying forests in urbanizing landscapes in Ohio, USA. We suspected that forests within more urban landscapes may be undervalued in our study system because (1) urban forests typically support lower densities of Neotropical migratory birds than rural forests and (2) anthropogenic disturbance and habitat alterations are likely to result in mismatches between cues typically used in habitat selection and actual habitat quality. In contrast to our predictions, field data suggest that urban forests are not undervalued. Our work not only expands upon previous habitat selection models by considering undervaluation, but also demonstrates how predictions derived from our model can be tested using a long-term empirical data set.