A spatially explicit optimal foraging model of Black-legged Kittiwake behavior based on prey density, travel distances, and colony size

We developed a spatially explicit population model (FORAGER) that simulates the foraging movements of Black-legged Kittiwakes ( Rissa tridactyla) in Prince William Sound (PWS), Alaska, an oceanographically, spatially, and temporally complex marine environment where the foraging behavior of this species has been extensively studied. Kittiwake foraging behavior had been previously modeled in the North Sea, an area spatially less complex than PWS, and we make comparisons to those studies. We calibrated and evaluated FORAGER by comparing predicted with observed bird movements, as well as with the predictions of a multiple regression model. Empirical data were derived from radio-telemetered kittiwakes tracked in an area where aerial surveys were determining the distribution and abundance of suitable prey schools over two years. The model simulated a predator searching for a school in a grid-based (2.78 km × 2.78 km) environment. The simulated predator is aware of fish-school density within a specified neighborhood around its current grid cell, and moves to an adjacent cell in order to maximize its prey encounter rate while avoiding cells that are primarily land. When a school is successfully found and exploited, the kittiwake returns to its colony by direct flight. Colony foraging grounds are created by simulating multiple foraging trips, each based on a different distribution of fish schools created by randomly perturbing the observed distributions. The four-parameter FORAGER model outperformed comparable multiple regression models, explaining on average 68% of the variation in observed kittiwake foraging behavior using a polar grid, and 44% using a rectangular grid. A four-parameter multiple regression model based on the same data set accounted for only 35 and 28% of the variation, respectively, depending on grid structure. Comparison of simulated and observed behavior suggests that kittiwakes in PWS plan their movements only within a neighborhood of about four grid cells, rely heavily on information gathered while foraging (including both negative and positive indications from conspecifics), and rely only somewhat on memory of prey distribution, as has been hypothesized to be effective in less complex regions (North Sea). FORAGER predicts that kittiwake movements should differ from the Ideal Free Distribution because of the predators’ uncertainty as to the actual prey distribution. Model results also indicate little relationship between the actual location o