Memory, not just perception, plays an important role in terrestrial mammalian migration

Abstract

One of the key questions regarding the underlying mechanisms of mammalian land migrations is how animals select where to go. Most studies assume perception of resources as the navigational mechanism. The possible role of memory that would allow forecasting conditions at distant locations and times based on information about environmental conditions from previous years has been little studied. We study migrating zebra in Botswana using an individual-based simulation model, where perceptually guided individuals use currently sensed resources at different perceptual ranges, while memory-guided individuals use long-term averages of past resources to forecast future conditions. We compare simulated individuals guided by perception or memory on resource landscapes of remotely sensed vegetation data to trajectories of GPS-tagged zebras. Our results show that memory provides a clear signal that best directs migrants to their destination compared to perception at even the largest perceptual ranges. Zebras modelled with memory arrived two to four times, or up to 100 km, closer to the migration destination than those using perception. We suggest that memory in addition to perception is important for directing ungulate migration. Furthermore, our findings are important for the conservation of migratory mammals, as memory informing direction suggests migration routes could be relatively inflexible.

Keywords: green wave; mammalian migration; memory; navigation; perception; zebra.

Figure 1.

The model domain showing the GPS data for zebras making the Okavango–Makgadikgadi migration in 2007 and 2008. Normalized difference vegetation index (NDVI) values (range: −0.15 to 0.93) are averages of 2007 and 2008 on 1 November at the start of the migration. Zebra data from [31]. (Online version in colour.)

Figure 2.

Distance from migration end point on the five day span when the GPS-tagged zebras ended their migration ((a) 27 November–1 December 2007 and (b) 15–19 November 2008). Random walk (RW), perceptual ranges (Inf indicates omniscience) and forecast horizon using past average conditions are compared. Outliers are omitted from the plot. (Online version in colour.)

Figure 3.

Locations of the modelled zebras at the time that the 2007 GPS-tagged zebras ended their migration. Contours show the 20th through 90th percentile using kernel density estimation. Migration start- and endpoints are marked with a triangle and circle, respectively. Perceptual ranges and forecast horizon using past average conditions are compared. Results for 2008 are broadly similar (see the electronic supplementary material, figure B4). (Online version in colour.)