Frequency-dependent foraging
To evaluate if pollinators exhibit positive frequency dependence in visitation (hypothesis 2), we asked if plant relative abundance predicted a metric that reflects pollinator preference (hereafter PI). We calculated PI for each species s at each sampling locationp as in (GrĂ¼ter et al., 2011):
\(\text{PI}\mathrm{s,p}=\frac{P\mathrm{\text{obs}}\mathrm{\ s,p}}{(P\mathrm{\text{obs}}\mathrm{\ s,p}\ +\ P\mathrm{\text{null\ }}\mathrm{s,p})}\),
where \(P\mathrm{\text{obs}}\mathrm{\ }\mathrm{s,p}\) is the proportion of pollen grains found on pollinators at site p that are speciess , and \(P\mathrm{\text{null}}\mathrm{\ }\mathrm{s,p}\) is the proportion of flowers of focal plants of species s among flowers of all plants in the transect. A PI value of 0 indicates no pollen from species s was found on pollinators at site p , a value of 0.5 indicates that the observed pollen amount matches expectations based on plant relative abundance, and \(\text{PI}\mathrm{s,p}\) approaches 1 if the amount of pollen from species s is much higher than that based on plant relative abundance. For these calculations we used a dataset that included plant species that were present in transects. We tested for a relationship between preference indices and plant relative abundance using linear mixed models with preference index as the dependent variable, relative abundance as the independent variable, and site and species as random effects.