Program and Agenda



Annual analysis of the North American Breeding Bird Survey is a large task, involving modeling and estimation of population trend for over 400 species of birds at multiple scales. Survey features such as range of years covers and numbers of missing years vary substantially across hundreds of strata. Models need to account for variation among observers, and need to be applicable even when species are not commonly detected Hierarchical models have provided the first real possibility for reasonable analysis of BBS data. These allow modeling of observer effects and spatial parameters as random effects, providing a rich modeling framework. These models provide the opportunity to explore alternative patterns of population change. For instance, models currently in use treat year effects on the logarithmic scale as normally distributed with a linear trend scale. An alternative model treats the first differences of these effects as mean zero random effects, allowing greater flexibility in pattern of population change, but typically requiring greater data resources. The problem is, how do we evaluate the adequacy of data for application of various models? How do we evaluate the adequacy of specific models for description of the data? How do we discriminate among models? Is it possible or desirable to have a single model to apply at multiple scales and across species? Answering these questions is made challenging by the complex nature of BBS data. In this talk we discuss the challenges presented by these questions and propose some possible solutions.


Link, W. A., USGS Patuxent Wildlife Research Center, USA,

Sauer, J. R., USGS Patuxent Wildlife Research Center, USA,


Oral presentation

Session #:S13
Date: 9/26/2014
Time: 17:15
Location: Emerald Mountain - Bible Point

Presentation is given by student: No