| Methods for characterizing patterns and behaviors in movement ecology can be simple, classifying movement strategy with net squared displacement, or complex, Bayesian partitioning of Markov models. However, estimating the timing of migration with any of these methods tends to result in inconsistent or difficult to interpret results. For instance, behavioral change point analysis is too sensitive and overestimates the number of states (migration or non-migration), while the more complex methods are too computationally intensive for this simple task. At the Nevada Department of Wildlife, we have deployed over 1600 collars in the last five years. To decrease the number of hours spent manually estimating migration, we developed an algorithm that uses binary search of the net squared displacement values. This method successfully identifies migration timing more often and with less human input than other methods. Here we present a brief overview of the algorithm as well as its applications to mule deer migrations in Nevada.
|