Predation risk driven by microhabitat structure restricts the habitat of small rodents in the Soutpansberg Mountains of South Africa

Background/Question/Methods

Live trapping of rodent species (Rhabdomys pumilio, Micaelamys namaquensis, and Lemniscomys rosalia) in the Soutpansberg Mountains of South Africa revealed strong preferences for grassy habitats whereas woodland habitats were startlingly depauperate.  This study endeavors to determine if predation risk can explain the observed distribution, and if the difference in predation risk is dictated by the proximity of woodlands or a change in microhabitat structure. Giving-up densities (GUDs) were used to measure predation risk along a series of four transects at three sites spanning the edge of woodlands and grasslands.  Transects consisted of 7 trays.  The center tray was placed at the grass edge and 2m separated neighboring trays except for the trays farthest from the edge which were placed 5m from their nearest tray.  Trays contained 40 sunflower seeds mixed with 1L of sand.  At sunrise and sunset, sand was sieved, remaining seeds were counted (the GUD), and trays were reset to allow rodent foraging.  The process was repeated for a minimum of five night periods.

Results/Conclusions

Proximity to woodlands may represent greater risk due to an increased encounter probability with woodland predators such as Genetta maculata.  Alternatively, woodlands contain sparse understory structure rendering rodents more visible than grassy areas.  If predation risk drives woodland avoidance, GUDs will increase along transects towards woodlands.  To study microhabitat structure, grassy cover was removed around trays at one random transect per site, and grassy cover was added to woodland trays at one random transect per site.  If microhabitat influences predation risk, GUDs will increase where cover was removed and decrease where cover was added. Analysis with GLM on SYSTAT 13 revealed a significant position effect (p<0.001).  GUDs increased along transects towards woodlands, and little foraging occurred at trays farthest in woodlands.  This supports the hypothesis that predation risk is responsible for woodland avoidance.  Likewise, GLM revealed a strong cover manipulation effect (p<0.001).  GUDs with added cover increased slightly, however GUDs where cover was removed increased to the point of being indistinguishable from unmanipulated interior forest GUDs.  This suggests the unusual distribution of small rodent species in the Soutpansberg Mountains is driven primarily by predation risk, and microhabitat structure is important in determining the magnitude of risk perceived by foraging rodents.