Counting Bees: How Mark-Recapture Methods Can Validate Wild Bee Sampling Protocols
By Emma Briggs
Wild bees are an ecologically important and incredibly diverse group of insects. With roughly 20,000 species described worldwide, they provide essential pollination services in a wide range of natural and agricultural environments. Unfortunately, though, wild bees are experiencing unprecedented population declines—likely from anthropogenic stressors such as pesticide exposure, habitat loss, introduced species (including non-native honey bees), and climate change.
With these declines comes a critical need to monitor bee populations. If we’re going to conserve wild bees, we need to know which species are declining (and how fast).
Bee-Sampling Blind Spots
Current wild-bee monitoring protocols incorporate multiple sampling methods, including aerial netting, pan traps (otherwise known as bee bowls), and vane traps. However, there has been much controversy surrounding the efficacy of these methods—see previous Entomology Today articles “Building a Better Bee Trap: Researchers Say Bee Bowls Are Overused” and “What’s the Best Method to Monitor Wild Bees?” just for two examples exploring this issue.
These sampling techniques only provide an index of bee population size. In other words, they yield a count of bees captured over a sampling period—not a count of all bees in a population. A challenge with index sampling methods is that you never know whether the bees you sampled represent, say, 1 percent of the population or 50 percent, and if that percentage is consistent across species and habitats. Indeed, it appears not to be.
For example, pan traps typically capture large numbers of small-bodied sweat bees but few large-bodied bumble bees and carpenter bees, and it seems unlikely that these capture rates are a good indicator of each species’ relative abundance. This uncertainty leads us to the crux of the bee monitoring debate: Countless researchers rely on aerial netting, pan traps, and vane traps to provide information about the abundance and composition of bee communities, but how do we know that their capture rates accurately reflect the underlying bee populations?
Mark-Recapture for Measuring “Detectability”
To address this question, our research team—from North Carolina State University, the North Carolina Wildlife Resources Commission (NCWRC), and the U.S. Geological Survey’s North Carolina Cooperative Fish and Wildlife Research Unit—conducted a mark-recapture experiment, the results of which we published this month in the journal Apidologie. Mark-recapture studies are different from traditional bee sampling methods, because they incorporate information not only about capture rates but also about detectability—or the probability that an individual in the population is actually observed during sampling. They do this by first marking a subset of bees in the population, then measuring the detection rates of the marked bees in subsequent sampling periods. With this information on detectability, mark-recapture studies can estimate the total number of bees in the population.
This ability to generate absolute population estimates is crucial, since we can use those estimates to ground-truth capture rates from traditional sampling methods. For example, if we compare mark-recapture population estimates to capture rates from an index sampling method like pan traps, we can determine if the traps provide an accurate representation of the abundance and species composition of that bee community. Although researchers have long suspected that traditional bee sampling methods are biased, this kind of comparison can take away the guesswork and begin to quantify those biases.
Our study took place in the summer of 2020 at three sites in the Butner-Falls of Neuse Game Lands near Raleigh, North Carolina. We selected these sites since our research partner, the NCWRC, was already conducting aerial netting and pan trap sampling for an ongoing study. Since the NCWRC was already conducting index sampling, it was the perfect opportunity for us to step in and compare its capture rates to mark-recapture estimates. For the mark-recapture study, we focused on four common bee taxa: eastern carpenter bees (Xylocopa virginica), leafcutting bees (Megachile spp.), two-spotted longhorn bees (Melissodes bimaculatus), and green sweat bees (Augochlorella spp., Augochlora pura, and Augochloropsis spp.).
The study was divided into five two-week sampling periods in which three days were dedicated to the mark-recapture study, followed by one day of sampling with pan traps and aerial netting by the NCWRC. During mark-recapture survey days, we captured bees at the sites using insect nets and transferred them into a “bee squeezer.” Once restrained, we marked the bees on the thorax with a dot of color using paint pens. We used a different paint color on each sampling day, and if a bee was captured on multiple sampling days, they could sport up to three different paint colors. Conducting field research during the start of the pandemic wasn’t easy—we faced bureaucratic delays, transportation issues, and supply shortages. But, with a little bit of luck (and a whole lot of effort), we managed to complete our fieldwork before access to our sites closed for hunting season.
Surprising Findings, But Questions Remain
After collecting and processing our data, we used linear regression to compare the absolute population estimates from the mark-recapture study with the capture rates from aerial netting and pan traps. What we found surprised us. Pan traps, which are widely used in bee monitoring studies, were not well correlated with the mark-recapture estimates. In contrast, aerial netting capture rates were well correlated, indicating that this sampling method likely provides a better representation of the abundance and species composition of the underlying bee community. While pan traps are still useful in bee diversity surveys, further research is needed to identify conditions where they act as a reliable indicator of abundance.
While studies comparing capture rates of index sampling methods to mark-recapture population estimates are common in other taxa (e.g., mammals, birds), we are the first (as far as we know) to attempt one for aerial netting and pan traps in wild bees. Additional mark-recapture studies are needed to assess other sampling methods such as vane traps or nest counts, especially at a multi-species community level. To be clear, we aren’t suggesting that mark-recapture methods replace all other bee sampling techniques—they are far too labor- and time-intensive for that. And they have their own set of assumptions that must be met in any study. However, mark-recapture methods do provide a feasible way to monitor selected bee species and to evaluate the performance of other potentially flawed sampling methods.