Bindoff, AD and Wotherspoon, SJ and Guinet, C and Hindell, MA, Twilight-free geolocation from noisy light data, Methods in Ecology and Evolution, 9, (5) pp. 1190-1198. ISSN 2041-210X (2018) [Refereed Article]
Solar geolocation is used to quantify the movements of animals tagged with sensors that record ambient light with respect to time. Global location sensor (GLS) tags are small, light, and present minimal drag or wing loading. They are affordable, and some can record data for several migratory cycles. These benefits mean they can be used in applications for which satellite tags are unsuitable. However, large errors in estimated locations can result if the sensor is obscured, especially around twilight, and sometimes the data obtained is unusable by existing methods of analysis due to this source of noise. This places limitations on the usefulness of solar geolocation in conservation and monitoring efforts.
All existing methods of analysis are dependent on twilights being identifiable or faithfully recorded. Instead, the method introduced here depends on the overall pattern of day and night to calculate the likelihoods for a Hidden Markov Model, where the hidden states are geographic locations. We call this a "twilight-free" method of light-based geolocation.
This method quickly estimates locations from otherwise unusable noisy light data. We use examples to show that the method produces tracks that are comparable in accuracy and precision to other geolocation methods. Furthermore, efficiency and replicability of estimated paths are improved because the user does not have to subjectively identify twilights. Other data sources, such as sea surface temperature and land or sea masks are easily incorporated, further improving location estimates and processing speed.
The twilight-free method offers new opportunities to researchers interested in the movements of animals that routinely have obscured sensors, or to analyse previously unusable noisy light data. It offers a fast, efficient, and replicable method for analysing tag data without the need for time-consuming pre- or post-processing. By increasing the yield of usable data from GLS tagging studies, researchers can more efficiently quantify where animals are going and when, and monitor changes in habitat. This is of fundamental importance to management and conservation efforts.