Subsampling analysis enables optimizing data quality and quantity decisions for biodiversity assessment

Background/Question/Methods

Ecologists and conservation practitioners must decide how much effort is feasible and necessary to assess community metrics, such as species richness. A related question concerns data quality: can easier-to-obtain field measures, such as abundance or richness of field-based morpho-groups, predict true species richness? Here we address both questions in the context of assessing a highly diverse group of insects, wild bee pollinators, across space and time. Our data set consists of 48 sampling events, in which we visited each of 12 sites four times. We collected both lower quality (floral counts, observational data on bee morpho-groups) and higher quality data (net-collected bee specimens subsequently identified to the species level in the lab). We used a subsampling simulation to determine (1) how well each of the field measures predicts species richness as measured by collected specimens, and (2) how much sampling effort is required to achieve a given predictive ability, as measured by the correlation between field measures and the specimen-based species richness across all sites.

Results/Conclusions

We collected nearly 1600 individual bees of 70 species, observed 2475 flower visits by 9 bee morpho-groups, and measured flower abundance and richness in 1920 1 m² quadrats. Data from some of our easier-to-obtain field measures, but not others, successfully predicted bee richness as assessed by specimen collection. Simple observation of total bee abundance correlated highly with specimen-based species richness (r=0.80). Furthermore, as little as one hour per site of abundance observations per site provided a correlation of 0.74 with richness across sites. In contrast, observational morpho-group richness was a weaker predictor of specimen-based richness (r=0.63). Richness of plants in bloom at a site predicted neither bee richness nor abundance. Counts of floral units were somewhat predictive of bee abundance (r=0.65), but are labor intensive and difficult to standardize. In sum, observing bee visits to flowers for a short time, rather than net-collecting and identifying floral visitors to species level, might sufficiently assess the richness of this highly diverse group while avoiding both destructive sampling and the intensive effort required for taxonomic identification.