Route selection in the North American Breeding Bird Survey is based on a quasi-stratified random sampling design motivated (in part) by the desire to achieve unbiased estimates of trends and other summaries of avian population status. In practice, some departure from design intentions is realized because active routes become concentrated around urban areas, and this yields unbalanced sampling with respect to habitat and land use patterns, and temporal changes in land use. The need to consider potential biases induced by factors not controlled for (or uncontrollable) by design has motivated the development of a model-based framework for conducting inference about population status and trend assessments from BBS data. The present modeling framework is sufficiently generic to allow consideration of designs that deviate from random sampling. Thus, for example, redundant information that results from clustering routes around urban areas, or targeted sampling to assess specific hypotheses (e.g., about the effect of land-use patterns on population status), can be viewed not as deficiencies in the design, but as features that necessitate extension of existing models used for assessment. In this paper, we consider whether the sampling design is relevant to conducting inference about population status and trends, and we provide a framework for addressing potential biases induced by an imbalance in spatial coverage of sampled routes.
Additional Publication Details
Spatial coverage and inference: Trade-offs between survey design and model assumptions in the North American Breeding Bird Survey
Wilson Ornithological Society and Association of Field Ornithologists Joint Meeting, April 21-24, Beltsville, Maryland. Abstracts