Fitting movement models to data: The case of the Serengeti wildebeest migration

Background/Question/Methods Many hypotheses have been proposed to explain the seasonal migration of the Serengeti wildebeest, but few studies have compared observed distribution patterns with environmental drivers.  Wildebeest census data collected monthly over a three-year period across the Serengeti ecosystem were used in conjunction with spatial data on rainfall, plant nutrients, surface water availability, and topography to determine which key resources explain the spatial distribution and migration of wildebeest.  Information-theoretic methods were used to fit 14 competing wildebeest movement models to observed distribution patterns using ordinary least squares and autoregressive methods at two spatial scales.  For the best overall model, we examined how the spatial extent to which wildebeest can perceive the landscape affects the migration.

Results/Conclusions Models based on the rate of intake and nitrogen (N) concentration of green grass provided the best model fits at both spatial scales tested, suggesting that digestive constraints and protein requirements may play key roles in driving migratory behaviour.  Tree cover and water availability improved model fits only marginally when added to forage intake and nutritional quality.  The emergence of a migration was dependent on the ability of the wildebeest to perceive changes in resource abundance at relatively large spatial scales (> 40-50 km) in the model.  When movement decisions were based solely on assessment of local conditions, the wildebeest failed to migrate across the ecosystem.  Our study highlights the potentially key role of strong and countervailing seasonally-driven rainfall and fertility gradients—a consistent feature of African savanna ecosystems—as drivers of long-distance seasonal migrations in ungulates.