Although the use of metabarcoding to identify taxa in DNA mixtures is widely approved, its reliability in quantifying taxon abundance is still the subject of debate. In this study we investigated the relationships between the amount of pollen grains in mock solutions and the abundance of highthroughput sequence reads and how the relationship was affected by the pollen counting methodology, the number of PCR cycles, the type of markers and plant species whose pollen grains have different characteristics. We found a significant positive relationship between the number of DNA sequences and the number of pollen grains in the mock solutions. However, better relationships were obtained with light microscopy as a pollen grain counting method compared with flow cytometry, with the chloroplastic trnL marker compared with ribosomal ITS1 and with 30 when compared with 25 or 35 PCR cycles. We provide a list of recommendations to improve pollen quantification. Environmental DNA metabarcoding is a molecular method that consists of investigating environmental DNA samples made of complex mixtures of genomes from numerous organisms 1. Due to new sequencing technologies and bioinformatics tools, metabarcoding has been increasingly used to identify taxa in environmental samples 1 to monitor biodiversity 2-4 , to investigate ecosystem functioning 5 and interaction networks 6-8 , in both aquatic and terrestrial ecosystems. Nevertheless, its reliability in quantitative approaches, which depend on the match between counts of high-throughput sequence reads and the amount of sampled biological material 2 , is still the subject of debate 9,10. While taxon identification can reveal individual diet breadth 11 , species richness, and the composition of habitats 2 , communities 12 and ecological networks 4 , taxon quantification provides knowledge on species evenness in those habitats, communities and diets or on the level of individual or species specialization in networks, all of which is very useful in ecological studies. Research on pollination and knowledge of the quantities of pollen transported by pollinators allow for the estimation of plant-pollinator interaction strength and hence it gives a more realistic representations of networks than those made possible using traditional approaches such as observing visits to plants by pollinators 9,13. Metabarcoding has been used in pollen studies to identify pollen in honey 14-16 , insect loads 6-8,17 , insect nests 18 , airborne samples 19 , and to quantify pollen abundance across various sample types. Several studies found significant positive relationships between pollen abundance (estimated using light microscopy) or pollen DNA quantities, and the abundance or the frequencies of high-throughput sequencing reads in experimental samples 10,16,17,20,21 , airborne samples 22,23 , insect pollen loads 21,24-27 or in brood cells of solitary bees 28. Conversely, other studies found low or no significant pollen-sequence abundance relationships when using ITS2 markers applied to pollen provision in...
