A bumblebee worker repeatedly forages on the same flowers, gathering nectar to bring back to her nest. This video illustrates the way her route evolves over time: repeated parts of the route are reinforced and grow brighter, while abandoned portions fade away. After around 60 trips, she has settled on a repeatable and efficient flightpath (although it’s not the best possible route).

 

Bees, and other animals that feed on renewable resources, like nectar, develop habitual foraging routes, known as traplines. People commonly describe a trapline route as one that revisits the same locations (flowers or flower patches, for bees) in the same sequence each time, and the idea is that this allows the animal to exploit predictable resources in the most efficient way. In reality, things are more complicated than that, and there is still a lot we don’t understand about trapline foraging.

 

I was interested in how these trapline routes develop over time, so James Makinson and I used harmonic radar technology to track every foraging flight made by individual bees, as they learned where they could find feeders and developed a trapline route.

 

To understand how each bee’s route changed over time, I developed a visualization inspired by the pheromone trails of ants. Individual ants lay scent trails as they go, allowing nestmates to follow their route and discover sources of food. The pheromones they deposit on the ground evaporate over time, unless they are reinforced by another ant laying down more scent. Because of this, the most popular paths, walked by many ants, develop strong odour trails and can be easily followed by new recruits, whereas the road less travelled gradually fade away. This helps the colony concentrate on the most profitable foraging areas and allows them to be flexible over time: if a food source dries up, disappointed ants will stop laying pheromones and what was once a highway will dwindle away, ensuring that new ants don’t waste their time following a path to nowhere.

 

This network of pheromone trails is a useful metaphor for thinking about the way the route of a single bee develops with time. At the start of her foraging career she must explore widely, sampling flowers in different locations. As she learns what’s available, she can forage more efficiently by flying between the best or most reliable flowers, pruning her network of possible routes util she is left with an efficient flight path.

 

In this video, we visualize the bee’s flight as a white-hot line she trails behind her as she flies. This metaphorical trail gradually cools and fades away, unless she flies back over the same area, in which case it is reinforced. As we watch over the course of about 60 foraging trips, we can see that her early, very convoluted routes gradually fade away, to be replaced by straight, efficient flights between feeders. Some segments of the route quickly stabilize: the initial leg, starting from the nest (at the bottom of the screen) and flying to the lower right hand feeder, and the stretch from the lower left to middle left feeders, are almost unchanged throughout. Other parts of the route continue to change and develop for longer, and despite the dogma that trapline routes stabilize onto a single, stereotyped sequence, the flightpath never becomes completely stable.

 

There is a clear pattern in which the early, length exploratory loops beyond the boundaries of our feeder array get largely eliminated, as do returns to empty feeders, but even after several days of refining her route, the bee still does these digressions from time to time. This makes sense when you recall that flowers are ephemeral resources: previously reliable flowers die and new ones are growing all the time. Thinking back to the ant trails, you want unprofitable branches of the trail to dwindle so that you don’t waste resources looking for food where there isn’t any, but you don’t necessarily want those branches to die out altogether: if conditions change, you want to be able to respond by changing your foraging patterns quickly, so keeping your information on alternative food sources and routs up-to-date is a wise strategy.

 

To make this video, I started with data from the harmonic radar which tell me the bees location, relative to the radar, once every three seconds (which is the time it takes the radar to scan through 360 degrees). First, I converted the bee position data into latitude and longitude coordinates, by triangulating her position using known landmarks in the landscape. The next step was to use Brownian Bride estimation to calculate a probability density function for the bee’s position between each set of radar coordinates. This allowed me to divide the landscape into ‘pixels’ – little squares, one meter on each side – and to calculate the probability that the bee flew through each pixel. I can visualize these probabilities as a heatmap, where each pixel is coloured in proportion to how likely it was that the bee flew through it.

 

Each frame of the video updates the heatmap to incorporate what the bee did next. As time goes by, the ‘probability’ value of each pixel goes down with each time step, so the colours will gradually fade away over several seconds of video. But the bee is constantly moving, so new values are always being generated, representing the probability that she has flown over each pixel of ground since the last frame of video was generated. These are combined with the decaying values from previous frames to come up with a new frame of video, each of which has a ‘heat’ value which summarises where the bee has flown, with a higher weighting given to recent movements, and lower weightings given to older movements. The entire thing is sped up (by around 500x) allowing us to watch two days of route development in less than a minute.

 

The route development of several other bees is shown below. You can see that, while certain trends remain constant (they switch from long, digressive exploratory flights to efficient, straight line movements between feeders; and they reduce the amount of back-tracking to empty feeders; but none of them ever fully stabilizes their route), each bee’s flight path evolves in a different way and they never end up with the exact same solution to the problem of how to fill up on nectar in the most efficient way.