Incorporating temporal heterogeneity in environmental conditions into a somatic growth model

Abstract

Evaluating environmental effects on fish growth can be challenging because environmental conditions may vary at relatively fine temporal scales compared with sampling occasions. Here we develop a Bayesian state-space growth model to evaluate effects of monthly environmental data on growth of fish that are observed less frequently (e.g., from mark–recapture data where time between captures can range from months to years). We assess effects of temperature, turbidity, food availability, flow variability, and trout abundance on subadult humpback chub (Gila cypha) growth in two rivers, the Colorado River (CR) and the Little Colorado River (LCR), and we use out-of-sample prediction to rank competing models. Environmental covariates explained a high proportion of the variation in growth in both rivers; however, the best growth models were river-specific and included either positive temperature and turbidity duration effects (CR) or positive temperature and food availability effects (LCR). Our approach to analyzing environmental controls on growth should be applicable in other systems where environmental data vary over relatively short time scales compared with animal observations.