Navigating the Landscape of Fear

Animals searching for food must navigate complex landscapes with varying terrain, food availability, predator activity, and shelter. Where and when should they gather food? To what extent should they engage in anti-predator behaviors such as vigilance or seeking refuge if a predator is detected? Optimal foraging theory (OFT) posits that animals balance potentially conflicting goals (such as feeding versus escaping predation) by making decisions that maximize some expected utility or reward. However, OFT models have generally considered highly simplified landscapes, either ignoring spatial variability or assuming that the habitat consists of discrete, internally uniform habitat patches. As a result, OFT has largely avoided the question of how animals should move from one potential feeding area to another, or between feeding areas and refuges.

We develop methods based on stochastic dynamic programming to find optimal foraging strategies, including optimal movement paths, in a continuous landscape with spatially varying predation risk. Our approach accounts for switching from foraging to escape behavior when pursued by a predator. Because contingent escape paths are considered for all visited locations, they influence the optimal foraging path even before threats are encountered. The optimal strategy thus depends on the animal’s level of hunger, the distribution of food, and the perceived threat distribution. The realized path is further influenced by actual predator encounters.

We illustrate our approach with two numerical examples: the first hypothetical with two food-abundant regions accessible only via high-risk areas, the second based on empirical studies on foraging Samango monkeys, Cercopithecus albogularis schwarzi . We find that the shape of the forager’s utility function (risk-averse, risk-neutral, or with state-dependent risk sensitivity) affects not only its choices of where to feed, but also the optimal paths to and from each feeding ground.

Our methods make it possible to compare properties of observed foraging trajectories with those predicted for different goal functions. Foraging trajectories can then provide additional information, along with other behavioral choices, about what quantity, if any, animals aim to optimize while foraging.