eDNA detects fewer bee pollinators than visual surveys and is influenced by floral morphology: A case study from Thailand

Environmental DNA (eDNA) metabarcoding offers a promising approach for monitoring pollinators and other flower-visiting insects, but its effectiveness for detecting key pollinators such as bees remains uncertain, particularly in tropical systems. We collected floral eDNA samples from five common plant species in Bangkok, Thailand, representing two floral morphologies (composite inflorescences vs. large tubular flowers). For each species, we conducted concurrent visual surveys and analyzed eDNA samples of both whole flowers and floral reproductive structures (anthers and stigmas) to (1) compare the taxa detected by eDNA versus visual surveys, (2) examine whether floral morphology influences pollinator detection, and (3) test whether sampling floral reproductive structures can help discriminate true pollinators from other floral visitors. Metabarcoding of the cytochrome c oxidase I (COI) gene from 30 samples yielded 196 insect ZOTUs across eight orders, compared to 21 taxa from five orders observed visually. eDNA frequently detected small, cryptic taxa (e.g., thrips, whiteflies, gall midges) missed in visual surveys, but detected only three bee species from six samples. In contrast, bees (9 taxa from 6 genera) comprised almost 80% of all floral visitors observed during visual surveys. Bee detection via eDNA was significantly higher in large tubular flowers than composite inflorescences, suggesting that more bee eDNA is deposited and retained in flowers when bees crawl inside corolla tubes during foraging. Anther–stigma samples yielded fewer insect taxa, and no bee detections, compared to whole flowers. Our results highlight both the strengths and current limitations of floral eDNA for pollinator monitoring in tropical environments. While eDNA can greatly enhance detection of small, inconspicuous floral visitors, current methods fail to capture the wide diversity of bee pollinators. Methodological improvements, including bee-targeted assays, are needed to fully realize eDNA’s potential for pollinator monitoring.