Extracting spatial networks from capture-recapture data reveals individual site fidelity patterns within a marine mammal's spatial range

Estimating the impacts of anthropogenic disturbances requires an understanding of the habitat-use patterns of individuals within a population. This is especially the case when disturbances are localized within a population's spatial range, as variation in habitat use within a population can drastically alter the distribution of impacts. Here, we illustrate the potential for multilevel binomial models to generate spatial networks from capture-recapture data, a common data source used in wildlife studies to monitor population dynamics and habitat use. These spatial networks capture which regions of a population's spatial distribution share similar/dissimilar individual usage patterns, and can be especially useful for detecting structured habitat use within the population's spatial range. Using simulations and 18 years of capture-recapture data from St. Lawrence Estuary (SLE) beluga, we show that this approach can successfully estimate the magnitude of similarities/dissimilarities in individual usage patterns across sectors, and identify sectors that share similar individual usage patterns that differ from other sectors, that is, structured habitat use. In the case of SLE beluga, this method identified multiple clusters of individuals, each preferentially using restricted areas within their summer range of the SLE. Multilevel binomial models can be effective at estimating spatial structure in habitat use within wildlife populations sampled by capture-recapture of individuals, and can be especially useful when sampling effort is not evenly distributed. Our finding of a structured habitat use within the SLE beluga summer range has direct implications for estimating individual exposures to localized stressors, such as underwater noise from shipping or other activities.

Automated summary

Understanding the habitat-use patterns of individuals within a population is crucial for estimating the impacts of anthropogenic disturbances. This study demonstrates the effectiveness of multilevel binomial models in generating spatial networks and estimating structured habitat use in St. Lawrence Estuary (SLE) beluga population using capture-recapture data.