Effects of seed mass on seed foraging behaviors of two sympatric rodent species

Throughout a semi-natural experiment, we investigated seed use by two
sympatric rodent species of different sizes, Apodemus flavicollis and A.
sylvaticus, focusing on how Quercus cerris acorn mass affected seed
predation, removal and dispersal by mice. A total of 20 adult A.
flavicollis and 25 adult A. sylvaticus were individually tested in an
open-air experimental enclosures. 40 Q. cerris acorns were collected in
the same forest where the rodents had been cought and were then weighed
with a digital balance (to the nearest 0.01 g) and individually marked
with a univocal alphanumeric code written on the seed shell surface. Each
test started when we placed the 40-acorn sample on the ground at
the center of the enclosure and lasted for three consecutive days. At the
end of the test, the final seed fate of each acorn or acorn leftover were
recorded and classified as: (1) eaten (partially or
completely) in situ (i.e., center of the enclosure), (2) not
eaten in situ, (3) removed and eaten (partially or completely), (4)
removed and not eaten (i.e., seeds that potentially could have been
successfully dispersed). Animals were released at their capture site at
the end of the tests, after three days of captivity. We used generalized
linear mixed models (GLMMs) with binomial distribution (family = binomial,
link = logit) to investigate the relationship between acorn fate
and acorn mass and to estimate the probability of a rodent specific
response pattern (i.e., foraging strategy) to seeds. The individual seed
was considered as a sampling unit and seed fate was modelled as a binomial
variable. We structured four models for foraging strategy and final acorn
fate: a predation model (binary data: 0 = not eaten, 1 = eaten), a removal
model (0 = not removed, 1 = removed), a site-specific predation model (0 =
eaten and not removed, 1 = eaten and removed), a dispersion model (0 =
removed and eaten, 1 = removed and not eaten). In the GLMMs, the
individual rodent was included as a random factor, rodent species as
categorical factor and acorn mass as covariate with an interaction term.
To evaluate interspecific differences in foraging strategies, we estimated
marginal trends from the GLMMs, enabling direct comparison of acorn mass
effects on seed fate between species while maintaining model integrity. To
account for potential nonlinear effects of seed mass on removal observed
in the raw data we included a quadratic term for acorn mass to test for
unimodal patterns in seed removal. Given the difference in body mass
between individuals of the two rodent species we tested, rodent body mass
was also included as a covariate in the removal model to account for
individual variation in handling ability. Finally, to assess interspecific
differences in the mass of removed acorns, we used linear mixed models
(LMMs), with the individual rodent included as a random factor and rodent
species as categorical factor. We applied the same model structure to test
for differences in the acorn mass to rodent body mass ratio associated
with removed acorns. We detected a species-specific response in
seed predation: A. flavicollis was more likely to predate heavier acorns,
while seed mass had no significant effect on the probability of seed
consumption for A. sylvaticus. Removal probability did not differ between
species suggesting similar removal patterns with respect to seed mass
across species, which were more likely to remove intermediate-sized seeds.
Apodemus flavicollis removed acorns with a seed mass to rodent mass ratio
significantly lower than the one observed for A.
sylvaticus. After removal, the likelihood of predation versus
that of dispersal resulted to be species-specific. Specifically, it did
not vary with seed mass for A. flavicollis, while A. sylvaticus tended to
disperse heavier acorns. These findings suggest that the two species might
play distinct roles in determining seed fate.