Large variation of suction sampling efficiency depending on arthropod groups, species traits, and habitat properties

Suction sampling is widely used to estimate arthropod abundance and diversity. To test the reliability of abundance data derived from suction sampling, we examined sampling efficiency across a wide range of arthropod groups and tested for effects of species traits, vegetation density, and differences between sites. Suction sampling efficiency was quantified by vacuuming an enclosed meadow area and subsequent removal of the turf, which was treated with heat extraction to collect the remaining arthropods. We obtained 250 pairs of suction and turf samples from seven grasslands with variable vegetation density. High suction sampling efficiencies between 49 and 86% were obtained for Auchenorrhyncha, Heteroptera, Araneida, Curculionoidea, Hymenoptera, and Diptera. In contrast, efficiencies were below 30% for Aphidae, Thysanoptera, Staphylinidae and other Coleoptera, and for soil arthropods such as Collembola, Isopoda, Diplopoda, and Formicidae. Efficiency varied significantly among habitats (sites) for most groups, often more than two-fold. Surprisingly, sampling efficiency for Hymenoptera, Diplopoda, and Collembola increased with vegetation density, probably because aboveground activity of these taxa was higher in dense vegetation. Suction sampling was nearly twice as efficient for spiders living in the vegetation than for spiders living near the soil surface, and cursorial and large-bodied spider species were more efficiently sampled than web-builders and small species. Depending on the sampling effort, suction sampling missed between 49% (one sample) and 31% (250 samples) of the spider species present. Suction sampling efficiency varied more strongly among sites and among arthropod groups than previously recognized. Abundance data derived from suction sampling are strongly underestimated, especially for arthropods living near the soil surface. Thus, comparisons of abundance and diversity between sites should be restricted to vegetation-dwelling species of the most efficiently sampled groups. The positive relationship of sampling efficiency with vegetation density demonstrates that variation in efficiency is mediated by arthropod behaviour.