Lessons from Georeferencing Bumble Bees

Off and on in the last year I have been documenting the distribution of the bumble bees (Hymenoptera: Apidae: Bombus) of North Carolina, which is a great state for looking at niche specialization in these insects. First, we have a well-curated and bee-rich collection here, due mostly to the efforts of Ted Mitchell, who wrote a two volume series to the bees of the Eastern US based on more than 30 years curating bees in the NCSU collection. Second, bumble bees are primarily a cool temperate lineage, and the climate of the southeastern U.S. is too warm for many of the species. From west to east, North Carolina offers habitats that span cool boreal habitats, the rolling hills of the near-sea level Piedmont, and the warm grassy coastal plain, plenty of geographic variability for revealing niche preferences.

I set about the task of studying these preferences by compiling georeferenced localities for the species of North Carolina from the NCSU insect database (most records) and other available museum databases (through GBIF) and mapping the points. I discovered that most Bombus species exhibit strong trends towards one or two of the three habitat regions of North Carolina, but there were outliers. Should we trust such outliers and move on, or check them in the collection? I decided to check them.

The first lesson I learned in this process is that geographic outliers did an excellent job of targeting misIDs – the majority of such outliers, in fact, were misidentifications. Concerned, I looked at the identifications of the rest of the specimens of a few of these species and discovered that these outliers made up most of the misidentified material. Using geographic distribution greatly helped to iteratively check the quality of these determinations.

bumble bee distributions in North Carolina
Distributions of Bombus vagans and B. perplexus in North Carolina, based on specimens in the NCSU Insect Museum and at GBIF. Photos by jlucier (left) and Greg Page (right).

The second lesson is that bumble bees are prone to misidentification. Bumble bee specimens are heavily georeferenced, likely due to their abundance in collections, their ease of species-level identifications, and the importance of documenting their distributional patterns given recent declines. While this makes them great for studying niche restrictive factors, bumble bees inherently can be prone to determination error.  The majority of bumble bee species are part of various Müllerian mimicry rings with other bumble bee species. People like to make quick identifications based on color alone but in most cases there are a couple species that are easily confused if one is not looking at the detailed morphology. One such common mistake in the eastern US is Bombus vagans and Bombus perplexus (see Figure), two species that have vague discrimination between them and persistently perplex us. The Bombus affinis drawer, for example, which is a species in strong decline and a target for multiple georeferencing projects, was rife with misIDs. Many of these could be explained by B. affinis being an “in-betweener” phenotype with color variation similar to several other species. This species also had some Bombus auricomus mistakenly placed among it, most likely an alphabetical error of curation.

Finally, I learned that georeferencing errors are not insignificant and should be also checked through mapping. Some of the niche outliers were errors in assigning localities, which can be tricky when two areas have similar names (see “the danger of using Google Earth/Maps to georeference“). Overall, I have come to appreciate the value of geographic data checking for georeferenced species data. Low error can be a big deal when trying to understand geographic restrictions at a local scale.

Tags: ,

One Response to “Lessons from Georeferencing Bumble Bees”

  1. Sam Droege says:

    Another difficult to ID, perhaps the most, is B. sandersoni, which appears to be increasing in the mountain regions, but is often/usually lumped with B. vagans in collections.

Leave a Reply