The value of twinned pollinator-pollen metabarcoding: bumblebee pollination service is weakly partitioned within a UK grassland community.

Predicting ecological impact of declining bumblebee (Bombus) populations requires better understanding of interactions between pollinator partitioning of floral resources and plant partitioning of pollinator resources. Here, we combine Cytochrome Oxidase 1 (CO1) barcoding for bumblebee identification and rbcL metabarcoding of pollen carried by bees in three species-rich UK pastures. CO1 barcoding assigned 272 bees to eight species, with 33 individuals belonging to the cryptic Bombus lucorum complex (16 B. lucorum and 17 B. cryptarum). Seasonal bias in capture rates varied by species, with B. pratorum found exclusively in June/July and B. pascuorum more abundant in August. Pollen metabarcoding coupled with PERMANOVA and NMDS analyses revealed all bees carried several local pollen species and evidence of pollen resource partitioning between some species pairings, with Bombus pratorum carrying the most divergent pollen load. There was no evidence of resource partitioning between the two cryptic species present, but significantly divergent capture rates concorded with previous suggestions of separation on the basis of foraging behaviour being shaped by local/temporal differences in climatic conditions. Considering the bee carriage profile of pollen species revealed no significant difference between the nine most widely carried plant species. However, there was a sharp, tipping point change in community pollen carriage across all three sites that occurred during the transition between late July and early August. This transition resulted in a strong divergence in community pollen carriage between the two seasonal periods in both years. We conclude that the combined use of pollen and bee barcoding offers several benefits for further study of plant-pollinator interactions at the landscape scale.