The ability to predict the trajectory of bird migration can allow people to take a set of preliminary, so-called conservation measures, including various actions to prepare the place of arrival of the flock to a certain place (adaptation of lighting, warning of the people, movement of potentially dangerous organisms for birds from places of possible stop, etc.)

Eastern curlew

For a mathematical description of such a problem, it is necessary that the model used can describe the probabilities of transition from one location to another, taking into account the fact that we do not directly follow the flight of birds, but only occasionally observe their location, and we do not always have accurate data. It is also necessary to take into account the time passing between observations. All these conditions are ideally suited to the choice of HSMM (Hidden Semi-Markov Model), which is structurally identical to HMM (Hidden Markov Model) but also takes into account the time between observations.

The researchers used the eBird bird movement observation dataset, collected not only by researchers but also by ordinary volunteers.

Estimated Eastern Curlew networks using 2019 eBird records. Edge widths depict the relative transition probabilities between nodes; node sizes represent relative sojourn time lengths. Dotted lines depict the node boundaries. Colors depict edges from the same origin.

The quality of the model was checked by measuring the error between the simulated and true values of the time duration of the flock stop, the direction of movement, and other data known from the dataset. The researchers also conducted a case study on data on the migration of the Far Eastern curlew between Asia and Australia.