Initializing the simulation and placing the objects on the map.
The figure above details the initialization of the simulation. The simulation always begins in the morning before foraging occurs, and the hive and bees can be seen at the center of the window. The purple flower patches are placed by searching through a resource density map from the forage mapping project, and instantiating new flower patches at points with very high density.
User parameters in the simulation window.
The program also contains a set of parameters that affect the environment as well as the honeybee behavior. The figure above lists all of the parameters that can be altered. For example, the “Probability of Following Dance” parameter can be changed at any point in the simulation, and it affects the probability that a foraging bee will follow another bee’s dance.
After starting the simulation, the honeybees begin foraging. They fill their crops with enough nectar to fuel their first foraging run, and slowly they begin scouting for flowers.
Honeybees focus on nearby flowers in the morning.
As is expected, the honeybees will find the flower patches close to their hive earlier. After harvesting nectar at a good resource, the bees will relay information about the location of that flower once they return to the hive. Other bees that decide to begin foraging have the option to either scout for new flowers or follow information that returning bees have provided. In the above animation, many scouting bees can be seen flying randomly around the map, but if you look closely you can see a cluster of bees flying straight toward the nearby flower patches. These bees are following waggle dances.
The nearby flowers are emptied of nectar; the bees must look for other resources.
Unfortunately, the four nearby flower patches don’t have an unlimited nectar supply, and the bees quickly deplete these flowers of food. In the animation above, many bees can be seen flying straight toward these flower patches, but the information they were given is no longer beneficial. At this point, returning bees will no longer dance about their resources. Past research suggests that bees that have decided to forage will become scouts if they can’t quickly be recruited by a waggle dance, which is why an influx of scouting bees is leaving the hive in the animation above.
The bees eventually find the more distant resources and start harvesting there.
If we zoom out, we can identify scouting bees that have found the more distant flowers. As they report back to the hive with new information, more honeybees leave the hive in search of these new, distant resources. Eventually enough information is present in the hive and the bees begin foraging at the new flower patches en masse. This is the benefit of foraging in a large colony. The spread of information through waggle dancing allows large colonies to make decentralized decisions about resource quality. If the nectar was depleted in the four nearby flower patches, an individual forager would likely spend all of its energy trying to find the distant resources. In a large colony, however, a single scouting bee can find distant resources and her information can benefit the entire colony. To better visualize the current state of the simulations, let’s look at more statistics.
The amount of foraging bees over time.
The graph above displays the amount of bees foraging during 12 hours of simulated time. The honeybees move once every “tick”, and in this case each tick corresponds to 9 seconds of simulated time (a day is 9600 ticks). In this case, foraging encompasses bees that are scouting, following dances, harvesting nectar, or returning to the hive. We can dive a bit deeper and compare the amount of bees scouting for new flowers versus the bees following waggle dances.
A comparison of bees following dances versus bees scouting for new flowers.
In the plot above, there is a peak in scouting bees in the morning as the bees start looking past the nearby flowers for better resources. This peak declines sharply as the scouts report back with information about new resources, and in the afternoon there is a steady influx of bees following dances. The peak in total foraging activity occurs at mid-day when bees are starting to be recruited toward the distant flowers. This peak occurs because in the simulation, bees are more likely to forage if other bees are performing waggle dances.
The accumulation of nectar in the hive over a 24 hour period.
Finally let’s look at a graph of the changes in hive nectar throughout the day. In the morning there is a sharp peak as honeybees transfer all of the nectar from the nearby flower patches into the hive. This peak quickly drops off as bees begin scouting for other resources further away from the hive, using a lot of energy in the process. Once the distant flower patches are found, a second peak in hive nectar occurs. Finally, the linear decline in nectar at the end of the day occurs as bees in the hive consume energy at a slower metabolic rate than they do while foraging.
So what? The simulation runs and the bees forage, but how does that benefit the Bee Lab? If the forage mapping project can produce accurate resource density maps on the scale of a few square kilometers, then these simulations will complete the puzzle. Let’s hypothetically place a honeybee hive out in a field. If we can map the actual floral resources around the hive, we can run the simulation and make predictions about foraging behavior in that particular location. If we also utilize the technology Gabe is developing to track both the bee foragers and accumulation of nectar in the hive, we can compare our simulated predictions with empirical data to help determine the accuracy of the simulation.
There are a few more features I plan to add before its accuracy can be compared with real data. In the context of this project, it is especially important that the flowers are effectively mapped from the real resource distribution maps to the simulation window. Currently, I have a very primitive method for adding flower patches to the simulation in locations where the resource maps indicate high flower density. Improving this method is now my first priority. The flower patches in the simulation detailed above have varying amounts of nectar, but they are approximated as point particles in the window. I am currently working on also giving flower patches a size and density. These properties will significantly affect how long and how quickly honeybees can forage at each patch. At this point, I will be assuming for each flower the amount of nectar and its concentration are constant, but if we are eventually able to identify different flower species with the forage mapping project then this parameter could be variable.
There is work still to be done, but the initial results of the simulation are promising. The honeybees are foraging and more importantly, they are making decentralized decisions about different resources via communication. If the simulation seems to accurately emulate real honeybee foraging, it can then be used to predict behavior in brand-new experiments to better refine our understanding of the complex foraging process.
Further Reading
Honeybee foraging behavior.
Schmickl, Thomas, and Karl Crailsheim (2004). "Costs of Environmental Fluctuations and Benefits of Dynamic Decentralized Foraging Decisions in Honey Bees." Adaptive Behavior 12: 263-177.
Anderson, Carl (2001). "The Adaptive Value of Inactive Foragers and the Scout-Recruit System in Honey Bee (Apis mellifera) Colonies." Behavioral Ecology 12: 111-119.
No comments:
Post a Comment