Plant and pollinator community effects
To determine if pollinators carry more conspecific pollen at sites with less heterogeneous plant communities (hypothesis 3), we characterized plant species richness, diversity, and evenness using custom R scripts (see data availability statement) and tested if site types differed in these metrics using t-tests. We tested if plant species richness, diversity, or evenness predicted the amount of conspecific pollen using generalized linear mixed models with a Poisson error structure and site as a random effect. To determine if pollinators carry more conspecific pollen at sites with more heterogeneous pollinator communities (hypothesis 4), we tested if insect species richness or diversity predicted the amount of conspecific pollen grains found on pollinators using generalized linear mixed models with a Poisson error structure and site as a random effect. To evaluate if the abundance of floral resources at a community level facilitates pollinator specialization (hypothesis 5), we quantified the total number of flowers found along each 90 m transect. We asked if the total number of flowers predicted conspecific pollen amount using a generalized linear mixed model with log-transformed flower number as the independent variable, site as a random effect, and a Poisson error structure.
Surprisingly, we found the opposite of what we expected regarding differences between urban and natural sites in the amount of conspecific pollen carried by pollinators (see results). Therefore, we also explored possible effects of the identities of individual plant and insect species on pollinator short-term specialization. We hypothesized that pollinators carry more conspecific pollen from invasive plants, which we predict to be in greater abundance in urban areas. To examine if plants’ invasive status predicted the amount of conspecific pollen we determined if each plant on which a pollinator was caught is invasive in California using the California Invasive Plant Council Dataset , as well as theCalflora Database. We determined if site types differed in the amount of invasive plants sampled along transects, using a generalized linear mixed model a binary dependent variable that reflected if a plant was invasive or not, site type as the independent variable, site and plant species as random effects, and a binomial error structure. We tested if invasive status predicted conspecific pollen amount used invasive status as a binary independent variable in a generalized linear mixed model with average conspecific pollen amount within sites as the dependent variable, site as a random effect, and a Poisson error structure.
To evaluate if differences in conspecific pollen amount were driven by differences in the identity of the insect species found among sites, we first determined which insects were common to all or the majority of sites. Most insect species were found at five or fewer sites, with the exception of one honey bee and one bumble bee species (Apis mellifera and Bombus vosnesenskii ), which were found at 10 and 11 sites, respectively. We hypothesized that if rare species were driving differences among site types, the change between urban and natural environments in the amount of conspecific pollen carried would be greater for common species. To test this hypothesis we modeled conspecific pollen amount as a function of the interaction between whether a pollinator was common or rare and site type, using a generalized linear mixed model with site as a random effect and a Poisson error structure.