Growth rates are intrinsically related to survival and lifetime reproductive success and hence, are key determinants of population growth. Discerning the patterns of growth in animals that exhibit complex life histories however, is a challenge. Marked shifts in growth can occur as a result of dietary shifts, different environmental conditions experienced across life-stages, or changes in resource allocation from growth to reproduction. Despite their importance in developing biologically realistic models however, such shifts are seldom accounted for in traditional growth analyses. Stage-dependent growth is omitted largely because data collected across life-stages are often incongruent; sampling schemes are chosen based on ease of detection, rather than ease of subsequent analysis.
Here we describe a hierarchical Bayesian growth model for the endangered salamander, Ambystoma bishopi. Annual mark-recapture data has been collected for the terrestrial adult stage, whereas aquatic larvae are sampled sporadically with no knowledge of individual identity; the combined dataset comprises 1492 capture events across six years. We modified the traditional von Bertalanffy growth equation to produce a single curve with parameters that relate to maximum growth rates in both larval and adults, and an additional two parameters that control the age and size at metamorphosis.
Results/Conclusions
The unified model provides strong support for two distinct growth trajectories in A. bishopi, commensurate with the species’ life history. Posterior distributions of estimated parameters indicate rapid larval growth, before reaching a plateau in size (45±2mm) coinciding with the transition to the adult stage. Following metamorphosis, growth resumes, albeit more slowly, and individuals approach a second, final asymptote (66±4mm). Combined, the two datasets provide complimentary information, with larval measurements shedding more insight on growth rates in younger age classes, whilst older individuals help to discern size at metamorphosis and the species’ maximum size. Initial growth rates for example, are drastically underestimated from the mark-recapture data alone. Thus we demonstrate that even with piecemeal data, the all-encompassing model is not only readily achievable, but also improves the accuracy of certain model parameters.
This model provides a crucial component in assessing the viability of remaining A. bishopi populations. More generally, the Bayesian methodology adopted here can readily incorporate multiple, disparate data sources, measurement error, and individual variation; as such it is ideally suited to many ecological problems. These tools have broad applications for modelling taxa with complex life-histories, and their utility is far-reaching given the ubiquitous nature of hodgepodge datasets in ecology.