Modeling Spatial Variation in Waterfowl Band-Recovery Data

Abstract

Historically, the statistical complexity of modeling spatial relationships in band-recovery data has limited the use of spatial models in the management of waterfowl populations. Consequently, managers have assumed simplified spatial relationships (e.g., by stratification and pooling data over large geographic areas) to obtain spatially explicit estimates of vital rates. As an alternative, we used a binomial random effects approach to modeling spatial variation in band-recovery data. The model accommodates spatial correlation and heterogeneity in recovery rates and facilitates spatially explicit estimation of recovery rates with sparse data and at arbitrary levels of spatial resolution. Although the model is structurally simple, estimation using conventional likelihood techniques is complex. Instead, we rely on a technique known as Markov chain Monte Carlo (MCMC) simulation. We used this model to construct a map of mallard (Anas platyrhynchos) recovery rates on a relatively fine-grained grid and for estimation of recovery rates within predefined geographic strata. The results show a strong gradient in recovery rate, with lower values in the western United States and higher values in the eastern United States. The spatial correlation in the model allows useful stratum-level estimates to be produced for strata with small sample sizes.