Relative role of dispersal dynamics and niche partitioning in among-individual variation in diet

Background/Question/Methods

The pervasiveness of among-individual variation in resource utilization has been demonstrated in many species and may potentially alter ecosystem function, population stability, and species interactions.   However, the causes of variation among individuals are still poorly understood. This study explores potential mechanisms for among-individual variation in diet to better understand its relationship to the ecosystem. Three alternate hypotheses were compared to determine if among-individual diet variation is due to dispersal from locations with different prey resources (H1, Dispersal), local resource partitioning (H2, Partitioning), or a combination of the two (H3, Full). These hypotheses were tested using carbon and nitrogen stable isotope data from eastern mosquitofish (Gambusia holbrooki) in the Florida Everglades.   Samples were collected from twenty-one sites representing gradients of food availability (edible periphyton and macroinvertebrate densities), hydroperiod, and fish densities.  Five metrics describing the variation in isotope data were calculated for each site.  Path analysis models were created for alternate hypotheses, based on isotope metrics and environmental data. Structural Equation Modeling (SEM) was used to compare models. A Bollen-Stine χ²analysis was used to determine model fit, and Akaike information criterion (AIC) was used to compare among valid models. Two model series were run.  Model series A included all macroinvertebrate species in a single variable in the SEM.  Model series B subdivided macroinvertebrates into four variables based on prey functional groups (detritivores, carnivores, omnivores, and herbivores).

Results/Conclusions

For Model series A, only the Full model passed the Bollen-Stine χ² test of model fit (p=.066).  The effects of food availability on niche variation were significant (β = -0.549, p <.001; β = 0.432, p = .003). For Model series B, all three models passed the Bollen-Stine χ² test of model fit (p ≥  .11, respectively). The Partitioning model had the lowest AIC (248) followed by the Full model (250) and then the Dispersal model (254).  However, the Full and Partitioning models have ΔAIC ≤  2, so both were considered as possible explanations for variation in diet.  In these models, the effect of relative herbivorous macroinvertebrate density on niche was significant (Partitioning: β = -0.551, p = .001; Full: β = -0.515, p = .003).This suggests that niche partitioning is the major factor contributing to the observed diet variation of G. holbrooki in this study.  Adding information about dispersal dynamics of these fish did not decrease model fit; therefore diet variation may also be influenced by dispersal dynamics.