Estimating survival of an anadromous salmonid using Bayesian hierarchical models applied to acoustic telemetry, biological, and environmental data

Hierarchical modelling is frequently used to model ecological processes because of its ability to handle complex ecological phenomena by decomposing them into naturally explainable sub-models. Hierarchical Bayesian approaches have gained widespread use in health, social, and environmental sciences, including in the estimation of demo-graphic parameters such as survival. In this study, we combine Bayesian hierarchical models with acoustic telemetry data to estimate survival probabilities for high-latitude populations of an anadromous salmonid, the Arctic char (Salvelinus alpinus), while in-corporating environmental and biological covariates to assess their impact on survival. The model we present here can also account for temporally varying detection probabilities due to changes to the acoustic receiver array design and seasonal variation in the detection probabilities related to environmental conditions (e.g., ice vs. no ice). As previously documented in this species, survival was high (> 0.87) and we found that the covariates pertaining to sea ice coverage and Fultons condition factor impacted the survival probabilities. Contrary to our expectations, high-condition fish had lower survival rates. Survival was also considerably lower during the summer (open-water) compared to winter (ice-covered) seasons. While the biological explanations and implications of these findings require further exploration, they nonetheless demonstrate the utility of this approach. Specifically, we present a hierarchical Bayesian model that can consider environmental and biological covariates while accounting for varying detection probabilities, a major concern of acoustic telemetry studies. The model can be easily adapted for other taxa with similar life histories where mark recapture data are available and can be extended to include additional environmental (e.g., salinity) and biological parameters (e.g., sex).