A Bayesian classification model to reconstruct lifetime movement patterns of riverine fish using environmental tracers

Abstract Environmental tracers, including both elemental concentrations and isotope ratios, are widely used to reconstruct the movement patterns of animals throughout landscapes. The methodology involves creating a map that describes the distribution of the environmental tracer across the landscape, an isoscape and then matching the values of the same tracer in the tissue of the animal (teeth, fish otolith, feathers) to determine provenance at one or more life stages. Classification models are commonly used to assign an individual to different areas of the isoscape. However, many of the current classification models are data intensive and may not account for (i) spatial autocorrelation (i.e. where an animal has moved is a function of where it was previously) inherent to data sets that use environmental tracers, (ii) species' movement ability which can influence region assignment or (iii) the propagation of errors from misallocation of locations early in the otolith time series. Here, we introduce a Bayesian classification model to estimate large‐scale movement patterns over the lifetime of freshwater fish that has relatively low data requirements, integrates spatial autocorrelation, offers an avenue to include movement capabilities and quantifies the uncertainty associated with the classification of fish movement throughout its life. We use a simulation study to test the accuracy of this model and then demonstrate functionality using a small otolith microchemistry data set (four species of fish collected at two sites) and a 87 Sr/ 86 Sr isoscape from the Mitchell River (Queensland, Australia) that accounts for spatial and temporal variation in water 87 Sr/ 86 Sr using water and mussel shell samples. The probabilistic framework of the Monte Carlo simulation allows uncertainty to be incorporated at each life stage, reducing the cumulative impact of misclassification and providing a more reliable reconstruction of lifetime movement patterns.