Alternative Definitions of Floral Resource Availability Alter Inferred Plant Importance in Plant-Pollinator Networks

Null models are widely used to infer pollinator preferences from plant-pollinator visitation networks, but these inferences depend on how floral resource availability is defined. This is particularly important for plants with complex floral structures, where individual flowers, floral units and larger inflorescences may each represent biologically plausible resource units. Here, we used Heracleum sphondylium (common hogweed), an Apiaceae species with hierarchical umbels, to test whether alternative definitions of floral resource availability alter null model inference in plant-pollinator networks. Plant-pollinator interaction data were collected from three field sites near Bristol, UK (Brandon Hill, Fenswood Farm and Leigh Woods), and hogweed floral abundance was quantified using three definitions: individual flowers, floral units and umbels. We found that the inferred importance of hogweed depended on the definition of floral resource availability. Hogweed was generally classified as a potential keystone plant when abundance was defined using floral units or umbels, whereas individual flower counts produced more variable classifications. Changing the resource definition of hogweed also altered the inferred importance of other plant species, with 12% to 39% of plant species × insect order classifications differing among resource definitions across sites. Broader inferred pollinator preference patterns also differed among resource definitions at some sites, although the magnitude of this effect varied among communities. These results show that floral resource definitions can influence ecological inference from the same observed visitation data. Explicitly defining and reporting floral resource units is therefore essential when using plant-pollinator networks to infer pollinator preferences and plant importance.