Bayesian Spatial Modeling of Data from Bird Surveys

• Activity Number: 269
• Type: Topic Contributed
• Date/Time: Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Raymond Webster*+ and Kenneth Pollock and Theodore Simons
• Companies: North Carolina State University and North Carolina State University and North Carolina State University
• Keywords: bird surveys ; count data ; CAR model ; detection probabilities ; capture-recapture

Abstract

Past analyses of large bird survey data have ignored variation in bird detection probabilities across space and spatial dependence in bird density. We present a unified framework for modeling bird survey data in the form of repeated counts, removal counts, or capture history data that accounts for both spatial dependence in density and variation in detection probabilities between survey sites. The hierarchical structure of the models makes them suited to Bayesian analysis using Markov chain Monte Carlo algorithms. For computational efficiency, we use a form of conditional autoregressive model proposed by Hrafnkelsson and Cressie (2003) for modeling spatial dependence. We apply our models to data from a large survey in the Great Smoky Mountains National Park. Our analyses have implications for survey design and show the limitations of fitting such complex models to sparse data.