Quantifying and reducing epistemic uncertainty of passive acoustic telemetry data from longitudinal aquatic systems

摘要

Passive acoustic telemetry data are used to study animal movement in aquatic environments, but tools to process the data are limited. In areas that are too large to be fully covered by the limited detection ranges of receivers or acoustic listening stations, researchers generally assume that animals are residing in an area when they are detected frequently at specific receivers. There is, however, no consensus on how this area and frequency should be spatially and temporally defined respectively, thereby introducing some unaccounted uncertainty of this residency-at-receivers. In longitudinal aquatic systems such as rivers or estuaries, strategically placed receivers are often used as gates or curtains through which tagged animals have to pass. Rather than being a proxy for the time spent near receivers, the detections can serve as boundary conditions to delineate the durations that animals spent between receivers (i.e. residency-between-receivers). As such, providing a spatial and temporal context to the detections and enabling the quantification of epistemic uncertainty. To assess the usefulness of this approach, we analyzed telemetry data for migrating eel in a longitudinal estuarine acoustic tracking network of 21 gates. Results revealed a logarithmic relationship between epistemic uncertainty and gate network resolution, which indicates that transferring information from the spatial to the temporal level has a positive effect on the epistemic uncertainty. The poor correlation between the at-receivers and the between-receivers approach indicates that the latter, although less precise, may be more accurate. It should be noted that the suggested approach assumes that the gates of receivers are perfect detectors. If tagged animals are able to pass these gates undetected the uncertainty will actually be higher than assumed. Our approach is therefore less suited for large open areas such as lagoons or seas where gates with high detection probabilities are logistically challenging. This approach allows the quantification and reduction of epistemic uncertainty by providing a spatiotemporal context to the detections. Since establishing and maintaining passive telemetry networks is generally expensive and the results these networks generate are often used in decision making, an assessment of the network quality and data uncertainty is vital.