Analysis of capture-recapture models with individual covariates using data augmentation.

Area of Science:

• Ecology
• Wildlife Biology
• Statistical Modeling

Background:

• Capture-recapture models are essential for estimating wildlife population sizes.
• Individual characteristics can influence an animal's probability of being detected.
• Existing methods may not fully account for these individual covariate effects.

Purpose of the Study:

• To develop and demonstrate a Bayesian framework for capture-recapture analysis with individual covariates.
• To facilitate the incorporation of continuous and discrete covariates into detection probability models.
• To provide a practical implementation using Markov chain Monte Carlo (MCMC) methods.

Main Methods:

• Bayesian analysis of the joint likelihood using a data augmentation scheme.
• Application of Markov chain Monte Carlo (MCMC) for parameter estimation.
• Implementation in freely available statistical software.

Main Results:

• The approach was successfully applied to meadow vole (Microtus pennsylvanicus) data, showing body mass as a covariate for detection.
• The model was also effective in an aerial waterfowl survey using a double-observer protocol, with cluster size as a covariate.
• Demonstrated flexibility in handling different types of covariates and survey designs.

Conclusions:

• The proposed Bayesian data augmentation method offers a flexible and accessible tool for analyzing capture-recapture data with individual covariates.
• This approach enhances the accuracy of population estimates by accounting for factors influencing detection probability.
• The practical implementation in open-source software promotes wider adoption in ecological research.