Does food-searching ability determine habitat selection? Foraging in sand of three species of gerbilline rodents

The distribution of three gerbil species in the Negev Desert, Israel differs in relation to sandy habitats. Gerbillus gerbillus occurs in sand only, Gerbillus dasyurus lives in all habitat types except sand, Gerbillus henleyi occurs in sand at high density periods only. We hypothesized that the reason for this distribution pattern is differential ability of species to forage in sand. We tested the ability of gerbils to search for seeds in sand layers of different depths and predicted that a) G. gerbillus will have equal search success at different sand depths, b) G. henleyi (which is twice smaller than G. gerbillus and G. dasyurus) will find seeds in thin layers better than in deep layers, and c) G. dasyurus will successfully find seeds in the thinnest layers only. We predicted also d) that G. gerbillus relies on olfaction for seed location, so its foraging success will be higher in searching for whole seeds than for seed kernels, whereas this will not be the case for G. henleyi and G. dasyurus. We examined the responses of the three species to odor of whole seeds vs seed kernels in Y-maze. In regards to seed depth, the search success of G. gerbillus was significantly higher when the whole seeds rather than kernels were offered in all treatments except the control, but sand depth did not influence the search success. The search success of G. dasyurus and G. henleyi did not depend on the type of seeds offered. The search success of G. dasyurus was lower in experimental (1, 3, 5 and 8 cm sand depth) than in control (1 mm sand depth) treatments, but did not differ among most experimental treatments. The search success of the smallest G. henleyi depended on sand depth for both whole seeds and kernels. Gerbillus gerbillus and G. dasyurus did not lose body masses in any treatment, whereas body mass changes of G. henleyi were influenced by the depth of sand in which gerbils foraged. No species demonstrated differences in response to whole seeds vs kernels in Y-maze tests.