Social waves in giant honeybees repel hornets

Abstract

Giant honeybees (Apis dorsata) nest in the open and have evolved a plethora of defence behaviors. Against predatory wasps, including hornets, they display highly coordinated Mexican wave-like cascades termed 'shimmering'. Shimmering starts at distinct spots on the nest surface and then spreads across the nest within a split second whereby hundreds of individual bees flip their abdomens upwards. However, so far it is not known whether prey and predator interact and if shimmering has anti-predatory significance. This article reports on the complex spatial and temporal patterns of interaction between Giant honeybee and hornet exemplified in 450 filmed episodes of two A. dorsata colonies and hornets (Vespa sp.). Detailed frame-by-frame analysis showed that shimmering elicits an avoidance response from the hornets showing a strong temporal correlation with the time course of shimmering. In turn, the strength and the rate of the bees' shimmering are modulated by the hornets' flight speed and proximity. The findings suggest that shimmering creates a 'shelter zone' of around 50 cm that prevents predatory wasps from foraging bees directly from the nest surface. Thus shimmering appears to be a key defence strategy that supports the Giant honeybees' open-nesting life-style.

Figure 1. The two different perspectives used in the experimental scenarios A and B.

The nest-specific axes were defined by their real-world coordinates: x, the horizontal (sideways) directions in regard to the vertical flat of the nest; y, the up- and downward directions; z, towards and off wards the nest as projected on the horizontal plane. The camera (cam) imaged the x-y projection in scenario A and the x-z projection in scenario B. Note, the distances of the hovering wasp from the bee nest (d_xz) were addressed in scenario B as projection on the x-z plane.

Figure 2

(A) Continuous assessment of abdominal thrust activity of the experimental Giant honeybee nest, while a predatory wasp was present in front of it (scenario A); the ordinate value gives the relative strength of thrust activity, the maximum value (1.0) refers to the maximum strength of shimmering (max W_peak) as observed during 300 s after the onset of waves; the threshold at rW = 0.4 discerns small-scale (blue area) from big-scale (red area) waves; (B) the time course of big-scale (red curve) and small-scale (blue curve) waves; curves show arithmetical means of the respective waves, thin vertical lines denote SEMs; abscissa, rel experimental time, time zero is defined by the onset of the abdominal thrust activity (corresponding to waving); (C) the rate of abdominal thrust activity per min of the experimental nest in two behavioral contexts, (C₁) ‘undisturbed by a hornet’ and (C₂) ‘disturbed by a hornet’; abscissa gives the categories of abdominal thrust activity as percentage of maximal waving strength; the blue columns refer to small-scale waves, the grey columns and the thin lines in the background give the respective rates of the opposite behavioral context for comparison. The distributions of the rates of abdominal thrust activities differed between C₁ and C₂ insofar that the proportions of the occurrences of waves regarding both states varied from one category to the other (Chi-square test, P<0.001, f = 49). Furthermore, under the presence of wasps the abdominal thrust activities show generally higher rates (Wilcoxon Signed Rank Test: P = 0.017).

Figure 3. The time courses of shimmering of Giant honeybees in response to approaching hornets (scenario B).

The waving strength W ( = number of abdomen-shaking bees per frame) depends on the hornets' distances from the nest d_xz (A) and on the hornets' flight velocities v_xz (B); time zero defines the onset of the waves; (A) five d_xz classes (Cd_xz = 1–5; coded in yellow to red; for definition, see Methods and Fig. 4,5) and (B) eight v_xz classes (Cv_xz = 1–8; coded in green to blue) of hornet flight episodes were considered; d_xz and v_xz class values were assessed from hornets in the 400 ms interval prior to the start of shimmering. Curves show arithmetical means, thin vertical lines denote SEM. For data details, see table 1 and 2.

Figure 4. The effect of flight behavior of predatory hornets on the waving strength of Giant honeybees (scenario B).

The waving strength W₆₀₀ gives the numbers of bees which had shaken their abdomens over 15 frames (definition see inset and text); it depends on the hornet's distance from the nest d_xz (A) and on the hornet's flight velocity v_xz (B); time zero in the insets defines the start of the shimmering waves (cf. Fig. 3); (A,C) red to yellow shaded areas define the five d_xz classes, and (B,D) green to blue shaded areas define the eight v_xz classes of hornet flights as used in Fig. 3 (for definition see Methods); open circles are arithmetical means, thin vertical and horizontal lines denote SEM; thick lines are regressions of the mean values regarding to 201 wave episodes with 317 flight episodes of two wasps (Cd_xz = 1 to 5; Cv_xz = 1 to 8); regression A: r = −0.969, P = 0.006; regression B: r = −0.963, P = 0.015. W₋₄₀₀, the amplitude of the waving strength 400 ms before the onset of the consecutive wave (see inset and text for definition), giving the residual shimmering strength under repetitive conditions; W₋₄₀₀ declines with d_xz (regression C: r = −0.989; P = 0.022), but inclines regarding v_xz at lower velocity levels (Cv_xz = 1–5: regression d₁: r = 0.975, P = 0.005) and shows an overall nonlinear relation for Cv_xz = 1–8 (regression d₂: r = 0.857; P = 0.027). All tests refer to Polynomial Regression (SigmaStat).

Figure 5. Effect of flight behavior of predatory hornets on the waving strength of Giant honeybees (scenario B).

Percentage of hornet episodes (n = 317) observed in the respective five d_xz classes (A) and eight v_xz classes (B) of hornets' flights; (C) the relationship between the flight velocities v_xz (abscissa) of individual hornets and their distances d_xz to the nest (ordinate); dashed lines (A,C) give the average hovering distance of the hornets (cf. Fig. 6). Regression of the means in C: d_xz = 0.215*v_xz+40.596; Cv_xz = 1 to 8; r = 0.962; 317 episodes; P<0.001); for definition and color coding, see Figs. 3,4 and Methods.

Figure 6. The hornets' responses to shimmering depend on the distance from the bee nest d_xz (scenario B).

Hornet behaviors were categorized in five distance classes Cd_xz = 1–5 (see inset, Figs. 3– 5 and Methods). Thick lines connect the arithmetical means, thin vertical lines denote SEM;. (A) Wasp behavior monitored for 1600 ms, starting 400 ms prior time 0, the onset of shimmering; Δd_{xz (1s)} values give the changes in the position of the hornet regarding its distance to the honeybee nest within 1 s after the onset of the wave (significance levels: *, P<0.05; **, P<0.01; ***, P<0.001; t-test); the horizontal dashed line, the average hovering distance (d_hov = 52.10±0.53 cm; n = 9757 images). (B) Correlation between Δd_{xz (1s)} and d_xz, the thick line gives the regression of means (Δd_{xz (1s)} = −0.449* d_xz+22.078; r = 0.998; Cd_xz = 1–5; P<0.001; 326 episodes); positive values of Δd_{xz (1s)} at Cd_xz = 1–3 represent movements of the hornets away from the bee nest and indicate avoidance responses; the response shown for Cd_xz = 4 is neutral, and the negative values of Δd_{xz (1s)} at Cd_xz = 5 outside dhov illustrate that the hornets usually approached the nest when shimmering started.

Figure 7. The hornets' behaviors during shimmering (scenario B).

The time courses of shimmering (A,D) and of the hornets' flights (B–C,E–F), ‘reactive’ (A–C) and ‘non-reactive’ (D–F) episodes (for definition, see text). Honeybee and hornet behaviors were synchronized to the start of the shimmering waves (abscissas give the time in milliseconds after the start of the shimmering waves). The hornets' behaviors are shown in terms of distance velocity vd_xz (B,E) and turning angle θ_xz (C,F); see inset and text for definition. Two classes of hornets were defined according to their distance to the nest in the 400 ms interval prior to shimmering: d_xz<45 cm (red circles, 84 ‘reactive’ episodes; 59 ‘non-reactive’ episodes), and d_xz>45 cm (yellow circles, 65 ‘reactive’ episodes; 108 ‘non-reactive’ episodes). Different brown-shaded areas define two test intervals in relation to the time course of shimmering (brown-shaded: 0–400 ms, grey-brown shaded: 400–1000 ms). Circles and bars give arithmetical means±SEM. Big full (red or yellow) circles give significant differences of the data in relation to the starting time of the wave at t = 0 ms (P<0.05; Holm-Sidak test, Friedman Repeated Measures ANOVA on Ranks).

Figure 8. The correlations of the waving strength with the flight behaviors of ‘reactive’ (A–D) and ‘non-reactive’ (E–H) hornets over the time course of shimmering (scenario B).

Abscissas, the waving strength W assessed by the number of abdomen-shaking bees per frame; ordinates, the hornets' behaviors measured by the parameter distance velocity v_dxz (A–B;E–F) and turning angle θ_xz (C–D;G–H) using the data of ‘reactive’(A–D) and ‘non-reactive’ (E-H) episodes (cf Fig. 7). Red circles and red lines refer to hornets which were near the nest (d_xz<45 cm) in the pre-wave period, yellow circles and yellow lines refer to hornets which were further away from the nest (d_xz>45 cm). Two time intervals were defined in relation to the time course of shimmering (0–400 ms: A,C,E,G; 400–1000 ms: B,D,F,H; for the coding of the brown shaded areas, see Fig. 7). Circles and bars give arithmetical means and their SEM; thick lines give the regression functions (test parameter, see Table S1), which refer to Multiple Linear Regression (Sigmastat) of the arithmetical means (see Methods).

Figure 9. ‘Confusion’ of hornets hovering in front of the experimental honeybee nest as observed in big-scale (red color code) and small-scale (blue color code) episodes in scenario A.

(A,B) Mean trajectories of the approaching hornet (for compilation of trajectories, see text); the sectors between the dashed lines define the four divisions of the mean pre-wave flight directions of the hornets for pooling the x- and y-values of the positions of the hornets approaching to the nest in 40 ms intervals. Thick lines, arithmetical means of x- and y-values of the hornet's position; horizontal and vertical bars, SEM. Note, that the trajectories before the onset of the waves (coded by black thick lines) are straighter than after the onset of the waves (coded by red and blue thick lines). (C) Hornet flight behavior under the influence of big-scale (full red circles) and small-scale (open blue circles) shimmering waves; ordinate, mD_xy (see inset for definition of the mean vector length mD_xy; grey segment, the defined turning range), calibrated between 0 and 1, the resulting scalar rel mD_xy is a measure of direction fidelity of the hovering wasp (circles and vertical bars, arithmetical means±SEM). The data show that the direction fidelity of predatory hornets is lowered by shimmering waves, by small-scale waves stronger than by big-scale waves; stars refer to significant (P<0.05, one-way ANOVA test) differences between the responses of the hornet to big-scale (n = 20) and small-scale (n = 33) waves per time interval. (D) The mean time courses of big-scale (red line) and small-scale (blue line) waves (cf. Fig. 2); ordinate, relative abdominal thrust activity (mean±SEM); abszissa, the relative experimental time (C,D) at time zero occurred the onset of shimmering; pre-wave sessions are coded by gray or black, shimmering sessions are coded by blue or red (A–D).

Figure 10. Hornets' responses regarding the flight velocity v_xy to big-scale (A,B) and small-scale (C,D) shimmering waves (scenario A).

Big-scale and small-scale waves were categorized according to a threshold of 40% of the maximal waving strength (cf. Fig. 2). The peak in flight velocity v_xy of big-scale wave episodes (A) from 240 to 360 ms after the onset of wave significantly differs from pre-wave v_xy values (*, P<0.05, One-way Repeated Measures ANOVA; 15 episodes). Small-scale waves (C,D) obviously do not affect the flight speed v_xy of the hornet. Note, that the acceleration pulse of the hornet (A) coincides with the time courses of big-scale waves (B). Abscissa, experimental time in ms; time zero defines the start of shimmering; lines connect arithmetical means, vertical bars give SEM. The sketches above the graphs symbolize big-scale waves (red-orange areas) as spreading over the nest, while small-scale waves (blue area) remain local processes.