Quantifying the effect of time on geolocation accuracy in seabirds

Light-based geolocation can be an effective tool for understanding movements and distributions of free-ranging seabirds, particularly during migrations and long-distance foraging trips. The light levels recorded by geolocators (global location sensors; GLS loggers) are used to infer latitude and longitude of individuals from day length and time of local midday relative to UTC (Coordinated Universal Time), respectively. However, there is an associated inaccuracy in these location estimates, associated with both systematic and random error. Previous studies have quantified this error by calculating distances to locations over the same time period derived from more accurate devices such as satellite transmitters (platform terminal transmitters; PTTs) or GPS devices. These approaches to quantifying error have focussed on twice daily or daily locations, whereas the aims of many studies using geolocation can be achieved by identifying areas used over long time periods, typically during the non-breeding season. We reanalysed data from a previous study where 12 Black-browed Albatrosses Thalassarche melanophris were tracked simultaneously with GLS loggers and PTTs. Rather than assessing location error over individual half-days or days, we took advantage of the principle of central tendency and calculated the distances between centroids obtained from GLS loggers and PTTs over different time periods. Our results show that overall geolocation error decreases with an increase in the number of days of data (Δ centroid distance: 76.5 ± 3.9 km [mean ± standard error] when using 30 d of location data), most likely due to a reduction in random error. Centroids are still subject to residual error, but researchers can have confidence that their accuracy is sufficient to answer many research questions for seabirds.

Details