A model of seasonal variation in somatic growth rates applied to two temperate turtle species

Modeling somatic growth of animals whose growth rates are seasonally variable is a challenge. Seasonal variation in growth reduces model fit and precision if not accounted for, and ad hoc adjustments to growth models may be biased or biologically unrealistic. We developed a growth phenology model (GPM) that uses a logistic function to model the cumulative proportion of total annual growth. We applied this model using two different approaches to datasets from temperate-climate populations of two freshwater turtle species that experience extended winter dormancy during which no growth occurs. The first dataset consisted of repeated intra-annual observations of sub-adult snapping turtles (Chelydra serpentina) tracked by radio telemetry, which we analyzed in a Bayesian context, focusing on growth over a single season. We then demonstrated a post hoc combination of the fitted GPM with a separate overall growth model. For the second application, we fully integrated the GPM into a hierarchical von Bertalanffy growth model, which we applied to a dataset of primarily inter-annual observations of juvenile midland painted turtles (Chrysemys pitta marginata). Specifying informative priors allowed us to fit the model despite the sparseness of intra-annual information in the data. We also demonstrate using the beta cumulative distribution function as an alternative to the logistic function in the GPM. We discuss incorporating prior knowledge about seasonal foraging and activity periods into growth models via a GPM as a transparent alternative to deterministic, implicit, a priori constructs.

Automated summary

Modeling seasonal growth variations in animals poses challenges, and the GPM model was developed to address this issue. Two different approaches were applied to datasets of freshwater turtle populations, highlighting the importance of Bayesian analysis and informative priors in fitting the growth model.