Linking acoustic telemetry data to spatial covariates in river networks with spatially explicit capture-recapture models

Spatially explicit capture-recapture (SECR) models extend classical capture-recapture models to include spatially-explicit animal locations and environmental covariates. SECR models have been widely employed in terrestrial studies to predict the population size and densities of animals assuming a closed population over a defined area. In this work, we extend and apply SECR models to a novel use-case that both uses a relative density formulation and accounts for biased tagging distributions to estimate parameters of habitat use from acoustic telemetry data in a branching river network.

Using SECR models, we predict how temperature distributions during the summer feeding season influenced how tagged Arctic grayling ( Thymallus arcticus ) distributed themselves through the Parsnip watershed in northcentral British Columbia, Canada. We found that the relative density of tagged Arctic grayling peaked at water temperatures of 12.4 °C. In warm years, relative densities were constricted as parts of the watershed became unfavorably warm. In cool years, fish were distributed widely throughout the watershed.

In acoustic telemetry, only the tagged population is available for detection. We highlight several specific considerations and assumptions for using this approach: i.e. (a) activity centres are assumed to remain in the same location throughout the study period, thus the time window of the study should be selected accordingly (e.g., exclude migratory periods); (b) inferences from acoustic telemetry data depict relative (not absolute) densities; and (c) spatial tagging effort must be defined in the model to ensure that predictions are not merely an artefact of tagging effort across space and time. When applied following these assumptions, this method is broadly useful for ecologists as it presents a quantitative way to merge automated telemetry datasets with discrete habitat parameters that drive population distributions through time and are relevant to managers and conservation professionals. Further, this method can be applied to branching river networks in which topological challenges have hindered other statistical approaches.