Blog Post

Rhonda Ward CS590 Final Project Blog

Honeybee Behavior

 

Proposal

Ethology is the study of the environmental and evolutionary effects on animal behavior. One effect on animal behavior is signaling and communication between one animal and another. One of the most fascinating and well documented animal behaviors is the waggle dance of the European honeybee Apis mellifera.  In the early 1900s Austrian researcher Karl von Frish and his team spent several decades observing the foraging behaviors of these honeybees. They determined that foraging bees communicated the location of a food source via a “waggle dance” that the foraging bee would perform upon its return to the hive. Both distance and location of the food source were communicated by the circular direction and angle of this dance. This symbolic language is simple to decipher if there is one foraging bee communicating the dance, but how do multiple food sources and multiple foragers affect the behavior of other bees in a hive?

This NetLogo model seeks to determine the effects of multiple food sources and multiple foragers on the behavior of follower bees in a hive. The model will begin with one foraging bee as the messenger of one food source location, which is then communicated to worker bees upon his return. Worker bees are then able to interpret and navigate directly to that food source. Once the basic mechanisms of the model are functional, then the programmer can manipulate the food source variable to determine the effect of multiple food sources on the behavior of the foraging bee. Additional tests can experiment with effects of having multiple foragers and their influence on the behaviors of follower bees. As a result, I propose to determine how the number of food sources and number of foragers affect the behavior of follower bees in a hive.

Executive Summary

This NetLogo model examines the communication and behavior of the European honeybee, Apis mellifera, as they forage for food. This behavior was extensively documented by Karl von Frish and his team in the early 1900s; they discovered that forager honeybees actually use a “waggle dance” to convey the distance and direction of nearby food sources. When a forager honeybee returns to its hive, it is surrounded by other worker bees that interpret and navigate the location of the food source based on this dance. The model is designed to test two variables on the effects of forager and follower bee behaviors. First, it examines the behavior of the foraging bee when given multiple food source options. If given a number of food sources of the same type and smell strength, then the foraging bee will choose the food source located closest to the hive because the foraging bee will use the least amount of energy to secure food. This model also tests the effects of employing multiple foragers in one hive. Follower bees are trained to surround and “listen” to the bee closest to them; if a hive utilizes more than one forager, then the follower bees will follow the path communicated by the first returning forager bee. If both hypotheses are correct, this, in turn, will result in the depletion of the closest food source. Under these circumstances, unique behavioral patterns should emerge as forager bees secure local food sources.   

 The NetLogo model utilized two breeds of agents, foragers and followers, who followed simple procedures according to their role in the hive. Forager bees were sent out on a random walk in order to secure a food location. They recorded the distance and direction (heading) of this source and reported it back to the hive upon return. Follower bees simply stayed at the hive and then followed a listen procedure to obtain the food location information from a forager bee upon his return.

In testing multiple food sources, the trials conducted did not conclusively determine a preference. Calculations were made to analyze mean, mode, and range of food distance, and only a small sample of range data points to the conclusion that the closest food sources were utilized by the forager bee. When examining heading, foragers visited the northern quadrant in a ratio of 2:1, but, again, with only forty trials, there is too little evidence and too many variables still out there for one to make a definitive conclusion. More research and programming needs to be conducted as far as time, original heading, and food source variables to make a clear determination.

When testing the effects of multiple foragers on which system a follower utilized, ten trials each were conducted with 2 foragers communicating to 10 followers and then 5 foragers to 10 followers. With multiple foragers, their headings and distances travelled were often very close, thus no clear preference was shown by who the followers listened to. Those trials that did render ranges in results show that follower bees actually listened to food location messages from foragers who travelled the greatest distances. As mentioned before, more research, programming, and trials must be conducted in order to eliminate other variables that could come into play, such as initial heading, forager return time, and mechanisms of the forager random walk. This model does serve as a good baseline of animal behavior study and, with time, could lead to new and interesting ways in honeybee foraging behavior.

 Report

Introduction

Behavioral ecology, or ethology, is the study of animal behavior. One area of study involves the signal and response among two or more animals. As environments change and resources become scarce, it is important to document how animals communicate in many areas, including reproduction, protection from predation, and food security. This NetLogo model examines the communication and behavior of the European honeybee, Apis mellifera, as they forage for food. This behavior was extensively documented by Karl von Frish and his team in the early 1900s; they discovered that forager honeybees actually use a “waggle dance” to convey the distance and direction of nearby food sources. When a forager honeybee returns to its hive, it is surrounded by other worker or bees that interpret and navigate the location of the food source based on this dance. This communication should be simple to decipher if there is one foraging bee sharing the dance, but how do the number of food sources and number of foragers affect the behavior of follower bees in the hive?

This NetLogo model is designed to test two variables on the effects of forager and follower bee behaviors. First, it examines the behavior of the foraging bee when given multiple food source options. If given a number of food sources of the same type and smell strength, then the foraging bee will choose the food source located closest to the hive because the foraging bee will use the least amount of energy to secure food. This model also tests the effects of employing multiple foragers in one hive. Follower bees are trained to surround and “listen” to the bee closest to them; if a hive utilizes more than one forager, then the follower bees will follow the path communicated by the first returning forager bee. If both hypotheses are correct, this, in turn, will result in the depletion of the closest food source. Under these circumstances, unique behavioral patterns should emerge as forager bees secure local food sources.    

Methods and Materials

For agent based models this section should include the specifications of your agents, environments and interactions.  It should also include the assumptions you are making.  Put your code snippets here as well.  Make sure it is well documented.

The honeybee model employs two breeds of agents: the initial food scout or “forager” and the worker bees at home or “followers.” Agents will own the variable of “food-location” that they remember as the source of food in the environment. The model begins with the hive positioned at a central location (0 0), with one foraging bee as the food source scout that employs a random walk as it travels from the hive. Once it finds a food source (a patch colored red), it records this location as its food-location variable and returns home. It relates this information to the follower bees upon its return, and followers then set their food-location variable the same as the forager’s. I haven’t decided whether to have them actually move to the food source or just report this in a monitor.

Once these basic mechanisms are functional, I can then test the food source variable to determine the effect of distance of multiple food sources on the behavior of the foraging bee. I will create a food-source slider in order to manipulate the number of food sources placed randomly in the environment. Each food source is a patch colored red to simulate a red cluster of flowers, and I will assume that each food source is equal in all ways: type, color, smell type and strength. Each food source will be assigned a number, and each number is also recorded as a turtles-own variable that can be reported in a histogram.

The next step utilizes additional tests to determine the effects of having multiple foragers on the behaviors of follower bees. I will keep all other aspects of the model the same and simply create a slider that manipulates the number of foragers sent forth from the hive. Forager number and food source number will be recorded on a monitor or histogram to determine which forager and which source were utilized. Qualitative observations will also be recorded on agent behaviors as they navigate through the environment. 

Part 1:

Independent variable: distance of food sources for forager bee

Dependent variable: food source chosen by forager bee (represented by identifying letter or number)

Part 2:

Independent variable: number of foragers

Dependent variable: forager and source utilized

Constants: type of food source (flower cluster type, color, size, smell etc), random walk of forager bee(s), initial behavior/characteristics of bees, size of world

Results

In experiment 1, one forager bee was presented with multiple food sources of varying distances and locations (headings). All other variables were held constant. Forty trials were run at a max of 500 iterations to determine the mean distance and the mean heading to see if any patterns emerged in forager behavior. Results are as follows:

One Forager with Multiple Food Sources

[run number]

[step]

mean [food-distance] of foragers

1

500

13.65183

3

500

12.14672

4

500

16.90298

2

500

12.82395

5

500

10.76232

6

500

14.56296

7

500

8.949197

8

500

19.88749

10

500

11.28155

9

500

17.45205

12

500

17.24838

11

500

0

13

500

14.79169

15

500

0

16

500

0

17

500

11.92683

14

500

15.48693

18

500

0

20

500

16.86677

19

500

13.37377

[run number]

[step]

mean [food-location] of foragers

 

13

500

350

 

20

500

314

 

10

500

59

 

11

500

149

 

3

500

46

 

7

500

0

 

4

500

135

 

14

500

11

 

6

500

354

 

16

500

51

 

8

500

259

 

18

500

61

 

12

500

0

 

2

500

0

 

17

500

71

 

9

500

137

 

5

500

328

 

19

500

247

 

15

500

338

 

1

500

300

 

In experiment 2, multiple foragers were created to test the effects of follower behavior. Which system or path is utilized when multiple foragers are used? Ten trials were run to examine which forager conveyed his information. These trials were run initially with 2 foragers and then another ten trials with 5 foragers. Location and heading information, as well as how many followers listened to each forager is listed as follows:

 2 foragers

Trial

Ticks

Forager 1 Heading, Distance & # of Followers who listened

Forager 0 Heading, Distance & # of Followers who listened

1

---forgot to record

307, 11.26

4 listens

310, 12.53

6 listens

2

1705

65, 25.80

9 listens

284, 10.24

1 listen

3

1156

302, 8.59

4 listens

285, 9.11

6 listens

4

4405

345, 14.81

6 listens

331, 11.32

4 listens

5

442

321, 11.66

4 listens

225, 13.53

6 listens

6

929

251, 13.21

7 listens

306, 12.78

3 listens

7

2091

276, 10.53

6 listens

195, 12.96

4 listens

8

302

327, 11.10

4 listens

267, 14.47

6 listens

9

141

107, 8.84

6 listens

320, 8.54

4 listens

10

479

73, 12.18

5 listens

265, 13.19

5 listens

Observations: Trial 1: heading and distance were very close together for both foragers; four followers listened to forager 1, while six followers listened to forager 0.

Trial 2: very large number of ticks. Overwhelmingly, nine followers listened to forager 1, whose distance was 2-1/2 times farther than forager 0. Forager 1’s heading was at 65 degrees, while forager 0’s heading was in a different quadrant, at 284.

Trial 3: listens, headings, and distances all very close.

Trial 4: very long time for foragers to return. Listens, headings, and distances all very close.

Trial 5: tick time 10 times quicker than trial 4. Listens, headings, and distances all very close.

Trial 6: heading and distances for both foragers nearly the same, yet 7 followers listened to forager 1, while only 3 followers listened to forager 0.

Trial 7: listens, headings, and distances all very close.

Trial 8: very quick tick time; listens, headings, and distances all very close.

Trial 9: quickest tick time. Forager distance to food source nearly identical, yet they went to different quadrants. Six followers listened to forager 1, while four listened to forager 0.

Trial 10: Low tick time, similar distances between foragers, though they went to different quadrants. Five listens each.

5 foragers

Trial

Ticks

Forager 1 Heading, Distance & # followers who listened

Forager 2 Heading, Distance & # followers who listened

Forager 3 Heading, Distance & # followers who listened

Forager 4 Heading, Distance & # followers who listened

Forager 0 Heading, Distance & # followers who listened

1

623

219, 8.43

3 listened

78, 18.39

4 listened

295, 10.31;      1 listened

205, 14.65;      2 listened

18, 10.42

0 listened

2

454

283, 13.04;      3 listened

72, 14.95,    3 listened

26, 11.43 0 listened

335, 14,46;      1 listened

309, 9.68 3 listened

3

2924

145, 16.50;    3 listened

63, 15.76;    0 listened

257, 13.89;     4 listened

266, 13,27;    0 listened

151, 16.50;    3 listened

4

535

198, 14.15;    2 listened

274, 12.81;    3 listened

209, 12.73;    1 listened

181, 13.69;    3 listened

301, 13.09,    1 listened

5

1002

179, 14.15;    3 listened

277, 19.15;    2 listened

8, 14.98;   2 listened

167, 14.10,    1 listened

169, 18.24;    2 listened

6

371

156, 10.21;      1 listened

78, 8.10;  1 listened

349, 10.39;       2 listened

127, 9.42;      2 listened

327, 11.72;      4 listened

7

1685

153, 12.20;    3 listened

330, 9.85;      0 listened

246, 16.14;    1 listened

258, 17.57;    3 listened

152, 14.44;    3 listened

8

2179

324, 15.85;    4 listened

23, 17.61;    0 listened

21, 17.99;     2 listened

236, 15.59;     2 listened

359, 15.68;    2 listened

9

3394

116, 14.30;    3 listened

2, 12.77;  1 listened

18, 15.63;    2 listened

113, 15.77;    3 listened

221, 15.73;    1 listened

10

1061

340, 12.43;    1 listened

132, 8.81;      0 listened

315, 17.15;    4 listened

108, 12.90;    3 listened

217, 8.69;       2 listened

Notes and Observations

Trial 1: forager was listened to the most, with 4 followers. He traveled the farthest distance. This is the second time in trials with multiple foragers in which forager who traveled the farthest communicated his results to the most followers.

Trial 2: am beginning to notice that forager distances are all about the same. This verifies my observations that show that they are landing on the closest red patch, which should result from my code. Often there are zones of no red patches, so bees spend a long time buzzing around searching for food, which is displayed in the long tick time.

Trial 3: four followers listened to forager 3, who had the median distance travelled. It is observed that different food location headings look like they may result from initial heading of foragers.

Trials 4-5: After inspecting and watching foragers, it appears that initial heading has an effect on food location if there is food in the forager’s path or to the right. With open, empty spaces, foragers roam farther distances in a right-turn wiggle motion in search of food.   

 Trial 6: four followers listened to forager 0, who travelled the farthest distance, though not by much.

Discussion

Experiment 1was conducted to determine if food location (as determined by distance and heading) had an effect on forager behavior. Forty trials were run at a max of 500 iterations to determine the mean distance and the mean heading to see if any patterns emerged in forager behavior.  The most obvious behavior I noticed was that, for some trials, the forager bee did not move at all; meaning, he kept the same heading and the same distance, in the respective experiments. I haven’t figured out what is causing this; I assume that this is a problem with my code, though I haven’t debugged this yet.

When examining mean distance of the food sources, the mode of twenty trials was 0, while the average was 11.4 steps. Because I felt that the 0 steps resulted from an error in code, I omitted those 4 trials, and calculated the average at 14.3, and there was no mode. The range between these distances was low, at 10.94, leading me to believe that the forager bee landed on flowers closest to its origin. What seemed to vary was the time it took the bee to find a flower. If time permits, further experiments should be run to determine the length of time it takes a forager bee to find a flower using a random walk. Further, I should spend more time researching the actual movement of honeybees: is it random or deliberate based on other variables (color, smell), and analyze how these have an effect. These are considerations to be researched and tested in the future.

When examining mean heading of the food sources, the mode again from twenty trials was 0, while the average heading was 160.5. Again, assuming that the 0 heading resulted from a bug in the code, I omitted the four trials delivering 0. After examining this data, I determined the areas that the forager visited and separated these into categories >100, 101-200, 201-300, and <300. I discovered that 6 times the forager went to a heading <100, visited the 101-200 area three times, 201-300 area two times, and <300 area for six times. Knowing that the forager always began at 0,0 but could randomly change direction, I found it inconclusive to determine a pattern in heading. There appears to be a preference to the northern quadrants; however, more tests must be conducted where initial heading does not randomly change upon setup.

I hypothesized that the flowers closest would be visited most frequently since it required the least amount of energy to locate. My code didn’t actually match this hypothesis, since I didn’t factor in an energy variable for the forager bee. This would require more time and coding expertise to examine. Despite this limitation, with the random walk, all methods of analysis (average and range) begin to point to the idea that the forager bee did indeed choose the flowers closest to the hive. I failed to hypothesize about heading, as that means of analysis didn’t arise until I started writing code; however, for future experimentation, I would require that initial heading stay static in order to analyze.

Experiment 2was conducted to determine how the number of foragers affected the behavior of the follower bees. Upon their return, which forager would the follower bees “listen” to, in regards to food distance and heading? Follower bees were examined after they listened to foragers upon return, to determine which forager was listened to. Trials began with two foragers, and then increased to 5 foragers as 10 followers listened.

Upon examination after 10 trials with 2 foragers and 10 followers, it is hard to determine which system is utilized by the followers. Eighty percent of the time, followers chose the foragers 50-50 or 60-40. Only two trials indicated that followers showed a preference. In trial 2, 9 out of 10 followers listened to forager 1 over the other. Interestingly, forager 1 traveled 2-1/2 times the distance as forager 0. His heading was 65, while forager 0’s heading was 284. Since both heading and distance were different for these foragers, I am unable to determine which variable had an effect. Only trial 6 showed any other variance in follower system preference, yet heading and distances for both foragers was nearly the same. As a result, I need to conduct many more trials to see if there is a meaningful difference. Otherwise, I am led to believe that, with a random walk and no difference in any other variables (flower cluster, time, and so on), there is no preference in distance or location as to who follower bees choose to listen to or follow.   

 Another 10 trials were conducted with 5 foragers and 10 followers. The data revealed no clear pattern or preference in who followers listen to. Two trials showed that the forager that travelled the farthest communicated to the most followers, an observation that arose with 2 foragers, yet I cannot explain why. Again, more trials need to be conducted to be able to effectively determine what, if any, effect multiple foragers have on who followers listen to. I was initially concerned that having the hive centered on the origin (0,0), followers would listen to only the first one who returned. However, I found the code to spread them out around the hive (layout-circle), so I avoided that problem. Now I need to spend more time and change my settings to zoom in, so that I can find out if it is the forager who first directly comes back to the hive that conveys his message first.  

Conclusions

 This model set out to determine effects on honeybee behavior as they foraged for food. First, I wanted to determine what effect the distance and heading of multiple food sources had on one forager’s food choice. I initially hypothesized that foragers would choose food sources closest to the hive, as that required the least amount of energy from the bee. I observed that, most often, forager bees did visit food sources closest to the hive; however, I feel that this behavior is a function of the random walk that I assigned to the forager bees. If more time permits, I need to factor in an energy variable to evaluate this more. Further experiments also should be done to determine if forager bee movement is actually random, or is it deterministic, as well as how time factors into the food source choices. In regard to food source heading, foragers visited the northern quadrants more frequently than the southern, though I cannot explain why. Perhaps this is a function of the initial heading or the random walk.

The second part of the experiment was to test how multiple foragers affect the behavior of follower bees back at the hive. I hypothesized that those who returned first would communication food location information first, and this would affect a greater number of followers. Ten trials were conducted with 2 foragers, and ten trials were conducted with 5 foragers, each time communicating to 10 followers. As the program code evolved, closer examination was made on the heading and distance of the food source locations rather than which forager returned home first. Some trials indicated that foragers who travelled farther distances communicated their message more frequently, though more trials need to be conducted to verify this. In addition, closer inspection of the hive and timing of the return must be analyzed to help determine which foragers convey their message more frequently. Currently, more variables must be tested and analyzed before a determination can be made as to which system is utilized.    

Personal Statement

I certainly did not achieve any major scientific breakthroughs with the honeybee model, yet it taught me a lot about analyzing animal behaviors, writing code, and disseminating results. So often in a scientific investigation, our results are inconclusive or our method doesn’t “work” or do what we expect it to do. This model is true in that sense, yet I learned a lot from the process. In programming the agents, I had to learn to think about what they are thinking and see things from their perspective in order to write my code. I spent the bulk of my time troubleshooting to get code to run or figuring out how to write code to make my agents do what I wanted them to do. Thus, I would say my biggest accomplishment was simply the experience and accomplishing the task. 

Acknowledgments

I would like to thank James Taylor for his valuable time and extensive knowledge as he guided me on this project and throughout the semester. I would also like to thank Irene Lee and all of the CS for All leaders, for granting me this opportunity. 

 

 

Comments

Permalink

Submitted by AnnNet Delaney on Fri, 05/10/2013 - 11:50

Bees have always fascinated me.  I am glad to see that you are doing a project on this subject.  It will be nice to see your code and how your project runs when you upload it.

Permalink

Profile picture for user rhondawm
Submitted by Rhonda Ward on Sat, 05/11/2013 - 02:50

Attached is the honeybee model as well as a pdf of the report, since the tables didn't format well on the blog post. 

Rhonda