1. Bescherming tegen roofdieren: Zelforganisatie van visscholen en vogelzwermen Charlotte K. Hemelrijk Behavioural Ecology and Self-organisation Centre for Ecological and Evolutionary Studies University of Groningen The Netherlands

2. Fish in schools Under attack of a shark, the school forms a vacuole around the predator

3. Starlings Starling display above the roost in Utrecht: shape is highly variable (Brodie 1976, Carere et al 2009)

4. Safety In Numbers  Neil and Cullen (1974):  Larger schools-> lower succes of attack by predator Cuttlefish SquidPikePerch School sizes 1 6 20 1 6 predators Capture/contact ratio

5. Advantages of grouping Protection against predation: – Dilution effect – Confusion effect – Early warning Energetics Finding food Finding mates How do individuals travel in groups?

6. ‚Self-organisation‘ models Cognitively, simple behavioural rules of individuals Complex patterns at a group level Self-organisation ‚Understanding by building‘ (Pfeifer & Scheier, 1999)

7. Click to edit the outline text format Second Outline Level  Third Outline Level Fourth Outline Level  Fifth Outline Level  Sixth Outline Level  Seventh Outline Level  Eighth Outline Level  Ninth Outline LevelClick to edit Master text styles  Second level  Third level  Fourth level » Fifth level Click to edit the outline text format Second Outline Level  Third Outline Level Fourth Outline Level  Fifth Outline Level  Sixth Outline Level  Seventh Outline Level  Eighth Outline Level  Ninth Outline LevelClick to edit Master text styles  Second level  Third level  Fourth level » Fifth level Models Of Fish Schools Attraction/ Individuals move at a certain speed and react to others depending on their distance: Robust Group-level phenomenon: Coordinated schooling Huth & Wissel 1992, 1994; Reuter & Breckling, 1994; Couzin et al, 2002; Hemelrijk & Hildenbrandt,2008 Avoid Align Move to t=1, t=1, t=2 Avoidance Alignment Attraction Blind Angle

8. Model of fish school Moving individuals follow three rules: Avoidance Alignment Attraction Hemelrijk & Hildenbrandt, 2008, Ethology

9. School shape: Adaptive? Lower detectability, because predators attack at front (Bumann, Krause, Rubenstein1997) How organised ? Intention or self-organisation? (Kunz & Hemelrijk, 2003; 2005; Hemelrijk & Hildenbrandt 2008) Oblong

10. Oblong Shape as a side - effect Two and three dimensional models, several group sizes, group compositions, two cruise speeds High Low Movement Length Width 0 50 100 200 300 400 500 → School Size 2 1.8 1.6 1.4 1.2 1 Length/Width (Kunz & Hemelrijk 2003; Hemelrijk & Kunz 2004; Hemelrijk & Hildenbrandt,2008)

11. Start ball-shaped school N = 600 26 24 22 20 18 16 14 12 Length Length, Width time (/s) Width Development of Shape Collision Avoidance → usually Slow Down & Move Inwards → Lengthening of Swarm

12. Supporting evidence Fast Hemelrijk & Hildenbrandt,2008; Kunz & Hemelrijk 2003 Density Larger schools shorter Nearest Neighbour Distance -> more frequent avoidance -> more oblong as a side-effect 0 50 100 200 300 400 500 → School Size 2 1.8 1.6 1.4 1.2 1 Length/Width 10 30 100 300 500 → School Size

13. Empirical data of Mullets Corresponds to the patterns of the model! Hemelrijk, Reinders, Hildenbrandt, Stamhuis (2010) Ethology Length / Width 10 20 30 40 50 60 # individuals Mean NND 10 20 30 40 50 60 # individuals 10 20 30 40 50 60 # individuals Stamhuis

14. Click to edit the outline text format Second Outline Level  Third Outline Level Fourth Outline Level  Fifth Outline Level  Sixth Outline Level  Seventh Outline Level  Eighth Outline Level  Ninth Outline LevelClick to edit Master text styles  Second level  Third level  Fourth level » Fifth level Click to edit the outline text format Second Outline Level  Third Outline Level Fourth Outline Level  Fifth Outline Level  Sixth Outline Level  Seventh Outline Level  Eighth Outline Level  Ninth Outline LevelClick to edit Master text styles  Second level  Third level  Fourth level » Fifth level Faster Schools Are 10 100 2000 School Size 1 0.995 0.99 0.985 Fast Polarisation Slow Turning Less  Are More aligned (real fish: Viscido et al 2004)  Less Collision Avoidance  Falling Back  Less Oblong 10 30 100 300 2000 → School Size Fast Length / Width 2.5 1.5 1.0 Slow Partridge et al, 1980, Saithe

15. Faster Schools: Denser Core, And Loser Tail, Because Less Fall Back Contrary to Breder (1959); Radakov (1973) 0.6 0.5 0.4 0.3 0.2 0.1 0 10 20 30 60 100 200 300 600 1000 2000 School Size Density #Individuals / BLU3 Core School Tail Speed and Density Fast Slow Densest 10% 25% backwards core

16. Oblong form Arises as a side-effect of coordination!  Due to falling back to avoid collision Question Why are bird flocks seldom oblong and why are they variable in shape?

17. Variable flock shape in all contexts Explanation: ‘telepathy’ (Selous, 1930) This talk: by self-organisation? Starlings above roost Starlings avoiding a predator Dunlins travelling

18. Our model of starling flocks, StarDisplay Flocking model with: 1. local coordination (attraction, alignment, avoidance) 2. few interaction partners (6-7, Ballerini et al 2008) 3. flying following simplified aerodynamics and banking while turning (Norberg, 1990) 4. attraction to the sleeping site (roost) (Carere et al 2009) (Hildenbrandt, Carere, Hemelrijk, 2010) Behavioural Ecology traits of birds, starlings also in fish model Hildenbrandt

19. Number of interaction partners (Ballerini et al 2008ab) Fixed (topological)  6 à 7 in Rome Interact by:  attraction  alignment  avoidance Part of flock Ri

20. Stay over the Sleeping Site Horizontal attraction Vertical attraction sleeping area

21. Fixed Wing Aerodynamics Balanced constant level flight Lift (L) Drag (D) Thrust (T) Weight (W) Speed (v) Lift and drag Flight force Update Velocity Location FF

22. Parameters From Starlings ParameterDescriptionDefault value Δu Reaction time50 ms v0 Cruise speed10 m/s = 36 km/h M Mass80 g CL/CD Lift-drag coefficient3.3 Lo Default lift0.78 N D0,T0 Default drag, default thrust0.24 N nc Number of interaction partners6.5 rh Radius of max. separation (“hard sphere”)0.2 m RRoost Radius fo Roosting Area150 m (Hildenbrandt, Carere, Hemelrijk, 2010, Behavioural ecology)

23. Resemblance to real flocks? Qualitative Quantitative

24. Model StarDisplay: Flocking manouevres by self-organisation shapes distance and angle to nearest neighbours, orientation of shape and density distribution but the flock-volume is smaller than real (Hildenbrandt, Carere, Hemelrijk, 2010, Behavioural ecology) Resembles real flocks in (10 events) (Ballerini et al 2008) : Model StarDisplay

25. Model Videos Qualitative similarity to empirical data of Rome (Hildenbrandt, Carere, Hemelrijk, 2010, Behavioural ecology)

26. Similarity to empirical data of Rome: Two Flocks Merge Video Model Very similar also in duration (1s between pictures)

27. Question Why are bird flocks seldom oblong and why are they variable in shape?

28. Measure shape of flocks Movement Length Width parallel to movement direction (L/W) Like in fish schools I1 I2 I3 aspect ratios, I3/I2 based on bounding box parallel to longest dimension Flocks and schools are flat (I1= thickness) Movement

29. Results Default situation sleeping area N = 2000 Trajectory of center of gravity of flock

30. Shape When Turning Sharply changes due to changes in volume due to different behaviour inside and outside roost N = 2000 sleeping area Banking (dg) Banking NND Volume (m3) Volume 0 10 20 30 40 50 60 -> Time Aspect ratio I3/I1 I2/I1 I3/I2

31. Flocks of rock doves Similar to model, volume changes during turning Nearest Neighbour Distance Pomeroy & Heppner 1992 Turning Time

32. Due to rolling while turning Starlings Lift (before rolling) Lift (after rolling) Gravity Centripetal force Centrifugal force Model without rolling: Rolling induces loss of altitude Altitude (m) Time

33. Loss of altitude during turns in Rock Doves and Steppe Eagles Altitude Turning Pomeroy & Heppner 1992; Gillies et al 2008 Rock doves

34. Flock size Small flocks have relatively smaller changes in volume due to more similar condition (above roost, or outside) more global interaction in flock Large (2000) Small (200) Time ->

35. Deviations of global velocity during movement approx. straightforward temporary sub flocks

36. Correlation length ξ (m) Deviation of Velocity Larger flocks less synchronised Larger groups have greater sub-flocks of similar velocity deviation like in real starlings (Cavagna et al 2010) Gradient in model, 0.44, in flocks of real starlings 0.35 Flock size L (m)

37. Larger flocks: weaker global polarisation Larger sub flocks differ in direction more flock shape is more variable global local polarisation polarisation 0 500 1500 2000 3000 4000 5000 ---> N

38. High # interaction partners (50) N=2000 causes stable shape due to more global interaction, stronger synchronisation Volume (m3) Time 6.5 int p 50 int p more static volume stronger polarisation

39. More interaction partners (50 vs 6.5) More polarised more ‘synchronised’ -> less variable shape Local 6.5 nb Global 6.5 nb local50 nb Global 50 nb Polarisation

40. Causes of changes of volume Different headings in flock, asynchrony (above roost, or outside it) no attraction to the roost no banking high # interaction partners time no attraction to the roost no banking high # interact.- partners default, 2000 inds N=200, small flock size (Hemelrijk, Hildenbrandt, PlosONE, 2011)

41. Oblong I3/I2 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 100 500 1000 2500 individuals L W Oblong in forward direction Length/Width movement direction Usually oblong in all kinds of directions, seldom in the movement direction Shape of flock when turning mildly

42. Turning of a flock Model StarDisplay Rock doves (Pomeroy & Heppner, 1992) Change of orientation of longest dimension of flock to the movement direction, repositioning of individuals in the flock: due to low variability of speed

43. Variability of speed Frequency -100 -60 0 60 100 Speed deviation (% cruise speed) N = 2000 birds, 10 m/s is low

44. Low variability of speed in fish model During turn, shape changes relative to movement direction, not oblong in movement direction

45. Turns and oblong shape in fish-model due to high variability of speed: individuals remain at ‘fixed’ positions t = 1 movement direction t = 5 t = 10 Speed deviation % cruise speed Slow Fast

46. Advantage of lower variability of speed Saving energy Confusing predators Avoiding collisions -> traffic

47. Conclusion: Variable shape of travelling flocks is a consequence of – larger flock size – fewer interacting neighbours – a heterogeneous environment – rolling-movement during turning – but not due to higher variability of speed Lower variability (adjustability) of speed induces Change of shape relative to movement direction Repositioning during turns Testable hypotheses for empirical studies

48. Predation Ultrasound attack by fish on herring

49. Predation? Model

50. Models of Fish Schools under attack Predator approaches Prey has two tendencies – Stay with the school – Move away from predator Herd SplitFlash expansion Predator

51. Starlings under attack Carere et al 2009

52. Model with bird-flocks under attack

53. Financial support European framework, StarFlag NWO pilot grant Startup grant from Rosalind Franklin Fellowship at University of Groningen

54. Starlings Flock shapes  Many flock types also in absence of predators  Singletons and small flocks more often where predation is low  Higher catch rate where predation is low Carere et al 2009