Predicting avian migration routes in 3D
Developing a mechanistic model to explain and predict animal migration patterns has long been a goal of migration ecologists. Modeling this natural process is difficult because of the limited biological tractability of many existing migration theories compounded by the mathematical complexity involved in describing existing migration datasets. Bottom-up hypothesis development has proven valuable for overcoming these shortcomings. Rather than ask what mechanisms are creating the data we see, this method uses scientific 1st principles to predict what an individual should do based on the laws of thermodynamics and then compares those predictions to natural data. This bottom-up approach applied to migratory animals suggests that individuals are physiologically constrained to an energy niche primarily determined by mass and, to a lesser degree, evolutionary history. Feedback loops both spatially and temporally shepherd individuals toward the physical location of their physiological optima at the mode of that energy niche distribution. Migrating individuals are moving spatially through time, in predictable ways, to continuously optimize their energetic efficiency. I hypothesize that an individual’s migratory routes are primarily determined by mass, complemented by a few shape and insolation parameters, rather than a suite of ecological drivers forcing selection for migration.
Migration is a difficult phenomenon to study because each ecological dimension has its own trajectory through time. The planet and its resources are spinning one direction while migrating animals and their resource-needs travel in another. Collapsing this complexity into something biologically tractable ultimately transforms our model into one where animal navigation acts as the response variable describing how an individual consumes and manages energy. My bottom-up predictive model operates via a suite of 4 sub-models that each achieve a scientific goals: (1) control for relative movement using a common simulation platform; (2) calculate a dispersal kernel restricting an individual’s movement to mechanically achievable distances; (3) further constrict that dispersal kernel with the thermodynamic limits of the individual; and (4) use Levy walk predictions to build a movement probability surface inside the remaining dispersal kernel. I then competed these predictions with available GPS data using likelihood methods to compare various parameterizations against each other. Results indicate that individuals with smaller thermal niches must let more earth move underneath them in order to stay in their niche. Because these individuals allow more earth to pass underneath them, they will show more pronounced migration patterns than individuals with larger energy envelopes.