DMS Statistics and Data Science Seminar

Speaker:  Ephraim Hanks (Penn State University)

Title: Joint modeling of individual migratory bird movements and species distribution patterns in space and time.

Abstract: Migratory and partially migratory species often exhibit significant variation in behavior, with individuals varying in whether or not they migrate, the distance they migrate, and the migratory path taken.  Understanding the variation in migratory behavior, as well as the full annual cycle of this behavior, is critical to conducting environmental impact assessments and managing the species.  The eBird citizen science initiative produces space-time species distribution maps of many migratory bird species, which provide insight into the overall distribution of the species over space and time, but cannot fully reveal variation in movement behavior.  Individual telemetry data give fine-scale information on variation in movement behavior, but are only available for a relatively few number of individual birds.  We develop methods to jointly model these two data streams, with models that scale from the individual level up to the population level.  We develop a novel model for full annual cycle migratory movement, and also propose a new framework for modeling time series of compositional data.  We develop fast numerical methods to allow us to fit these joint models to continental-scale migratory movement data.  We illustrate this approach through an analysis of Golden Eagles in the Western United States.