Science Insider
Making Sense of the Crowd
Automated tracking technologies allow researchers to monitor behaviors of all the honey bees in a hive simultaneously
Companies are getting better and better at targeted advertising. Every day, sophisticated computer algorithms track our online activity and location data in an effort to feed us ads that are more likely to stick. Although nobody is actually watching us, it can feel that way — like someone has glued targets on our backs.
For some honey bees, though, that concern may be valid, quite literally: Tagging workers is what researchers like Dr. Jacob Davidson do when they want to track activities of individual honey bees within a colony. Davidson is a postdoctoral fellow at the Max Planck Institute of Animal Behavior, Germany. He and his colleague Dr. Michael Smith use dabs of glue to stick unique QR codes to the thoraxes of thousands of bees, which can be read by a high-resolution camera pointed at an observation hive.
“Previously it just wasn’t possible to track many individuals over long time periods,” Davidson says. “Now, we can simultaneously track many bees over their entire lifetimes.” Video processing software keeps track of all the bees as they go about their routines, and the data acquired teaches us about their daily behaviors, task allocation, and interactions with each other. It’s a bit more tedious than using personalized cell phone data, but it works.
So much is still unknown about what actually goes on in a beehive, and this approach provides a window into life inside a hive. Indeed, a hive is akin to a black box full of Shrodinger’s bees1: The act of removing a frame, or even tagging bees, may change the very behavior being measured. Not to mention, gluing tags on the backs of thousands of bees is incredibly monotonous work.
Luckily, researchers at the Okinawa Institute for Science and Technology and the University of Cologne have developed a new method for “markerless tracking” of individual bees in an observation hive. In an article published in Nature Communications,2 they describe how they are able to continuously monitor the bees’ activities for as long as five months — almost an entire field season.
It turns out, with high enough resolution, bees don’t need their own QR codes. They have enough unique features for a camera to recognize them based on their visual appearance (along with other information), au naturel, also known as their “pixel personality.” Using this method, Dr. Katarzyna Bozek, who is also a professor at the University of Cologne, Germany, and study lead, was able to differentiate between dancing bees, idle bees, and bees inside cells, for example.
“Tracking is a long-standing problem and not much attention is given to systems like social insects,” says Bozek. She notes that other visual tracking applications usually focus on monitoring human crowds and street scenes, so existing software is better suited to that kind of image. In a beehive, though, objects are densely located, identical, and their motion is irregular, so it was a challenge to apply the technology to this system.
“What we found out is that by following bees, not tags, we get access to additional postural information,” Bozek explains. “We can, for example, tell if the bee just crawled inside of a honey comb cell or what is its body orientation. The behavioral repertoires that we extract from markerless tracking can be therefore more detailed.”
The technology is still being improved, and Bozek and her coauthors state that future possibilities include documenting sleep, trophallaxis, fanning, and scenting. Already, the method is yielding surprising discoveries. The number of bees who crawled inside cells spiked each and every night, for example, leading the authors to speculate that bees prefer sleep tucked inside, rather than on, the comb. “Not having to tag the entire colony allows us to scale up behavioral studies,” says Bozek. “It allows us to track a larger number of colonies without a massive increase in work load.”
But of course, there are limitations. The tracking software leverages additional information, like timing, the bee’s trajectory, and previous location to help narrow down the possibilities of exactly who could be where if that bee’s image has been temporarily obstructed. As a result, a bee cannot yet be identified again if she exits the hive and then returns, because this extra information has been lost and her appearance may change after foraging (she could be covered in pollen, for example).
Despite this, the possibilities are exciting. As the method’s precision and sensitivity increase, it could be used to learn more about varroa resistance behaviors, like grooming or varroa-sensitive hygiene, and even movement of varroa itself. Virus-infected bees could be introduced to the hive, and contact tracing combined with virus transmissibility could estimate how rapidly an infection may pervade the social group.
We could also finally test the hypothesis that the queen is ….
