Reinforced Learning under Competition Enables Partitioning of Resources in Foraging Bats

Every spring tens of thousands of female lesser long-nosed bats (Leptonycteris yerbabuenae) arrive pregnant to a maternal cave in the Sonoran Desert of Mexico after a long migration of more than 1000 km from central Mexico. During the lactation period, they rely on the nectar, pollen, and fruit of the Saguaro cacti (Carnegiea gigantea) as their main food source, while the Saguaro relies on these bats as its main pollinator. In order to reveal the foraging strategy of the lesser long-nosed bats, we used miniature GPS devices with an ultrasonic microphone to track bats` movement and behavior. We used a drone to create a 3D model of the visited cacti fields, characterized the cacti distribution and the number of open flowers. Analyzing bat movements in relation to their food distribution allowed us to identify visits to a specific cactus. We found that lesser long-nosed bats conduct long commutes every night, flying up to 103 km each way from the cave to the foraging site. They concentrate their feeding in a specific area inside the cacti field, visiting specific cacti very often thus maintaining a foraging home-range throughout the night and during consecutive nights. Using a model with minimum memory, we show that bats revisit cacti according to their previous experience. Our results demonstrate a simple efficient strategy for foraging under competition uncertainty.