The dangers of data bias: a study on bees

What is it about?

After three and a bit years, the PREDICTS project (www.predicts.org.uk), with the help of its many data contributors, have amassed a large, taxonomically and geographically representative database of biodiversity facing different land-use pressures. This is a remarkable achievement; for most fields in ecology, we’re still well off the mark. In pollination ecology in particular, unrepresentative data has been recognised as an issue; ecological data on bees are often most readily available in North America and Western Europe, a geographic bias that also leads to a taxonomic one, as bumblebees are common in these areas. This is potentially a big problem; if different regions and taxa show different responses to land use, then biases in the underlying data can lead to misleading predictions when we try to generalize models to future impacts or to other regions and taxa. However, until recently, it wasn’t clear whether these biases in data really affected our inferences. In this paper, we explored this question. As suspected, we found that bee communities respond differently to land-use impacts depending on the region. This is in part because species in different regions may vary in how sensitive they are to human impacts (with bumblebees often responding differently to other bees) and also potentially because the nature of threats vary regionally. These results suggest that global extrapolation of models based on geographically and taxonomically restricted data may underestimate the true uncertainty, increasing the risk of ecological surprises.

Featured Image