In addition to light, the use of pressure and activity data recorded by multi-sensor geolocators has recently been shown to improve the estimation of a bird’s position. At the same time, modelling a bird’s trajectory with an MCMC sampler becomes more challenging when integrating this new information.
In this work, we propose to model the trajectory of a bird with a graphical model, allowing to compute and efficiently store the probability distribution of the entire trajectory. We demonstrate how the graph representation can be used to compute the following products: (1) the most likely path, (2) the probability map for each stationary period, and (3) simulated paths. This method is applied to 16 tracks from 9 different species.
The graph approach is mathematically exact (i.e., not iterative/approximation used) and relatively fast to compute. The trajectories produced combine information of light, pressure, accelerometer, and wind data, resulting in the highest spatial precision ever achieved for geolocator data. The increase of precision enables the reconstruction of the full journey of a bird, including all stopover (even only hours long). In addition, this approach also allows to retrieve airspeed and windspeed for each flight allowing further analysis on migration energetics and wind use.
The method presented here, allows for a reconstruction of full migratory pathways from geolocators at an unprecedented spatiotemporal resolution. Combing behavioural activity during stopovers with information on relative energy expenditure during flight bouts can reveal further aspects of stopover habitat quality, as well as potential carry-over effects, and by including breeding success, even fitness consequences.