Modeling sighting heterogeneity and abundance in spatially replicated multiple‐observer surveys

Failure to account for dependencies among observers in multiple‐observer capture‐recapture studies will lead to biased inference yet methods that account for dependencies are poorly developed. We combined hierarchical capture‐recapture models and finite sampling theory to infer population abundance of koala (Phascolarctos cinereus) in the 5,812‐ha Mount Eccles National Park, southeastern Australia. We used plot replication to build a Bayesian hierarchical model of koala sightings by multiple observers. Given the number of koala present on a plot, we modeled counts of koala by a multinomial log‐linear model with interaction terms to account for between‐observer dependencies. We modeled total numbers using a Poisson log‐linear model with covariate and random effects. We estimated the number of koala as 1.9 koala/ha (95% CI = 1.6, 2.4). We found evidence of between‐observer interactions, highlighting the importance of having at least 3 observers in multiple observer studies.