Impact of hornet density on bees and hornets flight performance
The speed of honey bees leaving the hive was negatively affected by the number of hornets present in front of the hive (LM, F=4.617, p=0.032). However, the speed (LM, F=19.36, p<0.001) and the curvature of the trajectories of honey bees entering the hive were positively affected by the number of Asian hornets in front of the hive (LM, F=59.74, p<0.001) (Figure 5 ). The number of hornets in front of the hive strongly reduced the variance of all parameters (speed, curvature and percentage of hovering) of bees entering, leaving the hive, and hornets (Figure 6), but only tend to decrease (no significant statistic) the variance of flight curvature of bees entering the hives (Table 2, Figure S3, Table S2 ).

Discussion

The automatic selection of scenes of interest is a significant step forward for behavioural ecology to reduce the manual labour involved. In this study it reduced 90 observation hours into a few hours, and allowed us to switch from 603,259 registered trajectories to 5,175 scenes of interest (i.e. short video segments) where both prey and predator were present. The field of view was appropriately selected, as it included the honey bees’ speed deceleration area as they enter their colony, through its reduced entrance, that is strategically exploited by predating hornets. Such data enabled the analysis of flight performance characteristics, namely speed, curvature, and hovering (established via 3D accuracy), of honey bees and hornets, to better understand their prey-predator interaction and look for potential drivers of predation success.
The flight speed of honey bees arriving and entering the colony was half as fast as the honey bees leaving the colony, probably because of the need of honey bees to slow down in vicinity of the nest entrance to correct their flight and enter the nest safely, i.e. “parachute” behaviour. To illustrate the importance of such parameter on the potential predation capacities of predators on bees, we can cite the very particular “crashing behaviour” of some Melipona bee species. In order to prevent predation in front of their nests by spiders, while they attempt to enter their earth built nest, these bees have adapted the shape of the nest entrance into a tube with a flat platform orientated in a specific way to allow these bees to bounce on it to enter the nest quickly, without having to decrease speed, therefore being harder to catch by predators waiting at their nest (Shackleton et al., 2019). Concerning flight curvature, honey bees leaving the hive had very straight trajectories, while honey bees entering their hive had more curved trajectories. The time spent hovering by honey bees leaving the colony was very low. Hovering seems to be a rare behaviour in honey bees, linked with specific situations (e.g. traffic jams), while in hornets this is characteristic of their predation flight in the hive vicinity or in other “hunting sites” e.g. at other insect aggregations (bins, animal carcasses).
The flight speed of honey bees in front of the hive increased in the morning, then for honey bees leaving their hive we observed a slowing around noon, to gently decreasing speed through the afternoon. It could be connected with a “traffic jam” phenomenon: when honey bee foragers departing the hive reach a maximum at the same time when young honey bees exhibit their first orientation flights outside the hive (Capaldi & Dyer, 1999; Capaldi et al., 2000). Hovering flights were mostly observed with hornets and returning honey bees. Hornet hovering decreased through the morning until early afternoon, and thereafter increased. This might be related to the quantity of potential prey available around the colony entrance: hornets may have to hover, to wait for predation opportunities. For the honey bees entering the hive, we observed a similar pattern, but with a slight increase in hovering in the early afternoon, potentially linked with the same traffic jam phenomenon described above. This hovering activity of bees in front of the beehive could provide hornets with more opportunity to prey on them.
During the course of the day, the number of honey bees leaving their hives and hornets in front of it at any one time follows a quadratic pattern, meaning that it is enhanced through the morning up to the middle of the day, to then slowly decrease at the end of the day. The optimum period of activity for hornets was around 12pm, while for honey bees peak activity around 2pm. This interesting offset could be linked with the predation success of hornets that decreases when there are too many hornets hovering in from of the hive. Arriving early at the hive would, even before the prey are abundant would limit the number of predating competitors. The presence and the temporal dynamic of honey bees in front of hive was very similar to what observed in Struye et al. (1994). The daily dynamic of the presence of hornets in front of the hive also matches with former works studying the rate at which hornets were leaving their nests over a day (Monceau et al. 2017; Poidatz et al. 2018).
Predator capture rates are expected to depend on encounter probability with prey, prey escape capability, and on predator agility (Kruse et al., 2008). We confirm that hornets succeed in preferentially catching honey bees going back into their colony (Shah and Shah, 1991): we first hypothesised that the main explaining parameter could be their speed, which is lower than for bees leaving their colony, as they come back more slowly due to their nectar-filled crops or pollen baskets, and need to slow down to be able to access the small hive entrance. But our analysis showed that, whilst returning bees were slower, the flight speed did not play a significant role in the probability of success of hornet predation. We also reveal in this study that the quadratic number of Asian hornets in front of the honey bee hive influences their predation success on honey bees, reaching a peak at 8 hornets, above which threshold their predation success decreases. In the increasing phase, there are enough active prey, and we can hypothesise that the predators are distributed at the hive entrance in a way that allows them to optimize space and occupy more and more potential honey bee paths; but above 8 hornets, hornet predation success decreases with their increasing number. It could be due to inter-specific competition (Monceau et al., 2014), and maybe also be due to the foraging paralysis of honey bees, that do not exit the colony at the same rate anymore.
This result is congruent with the results of the escape success in terrestrial predator-prey interactions model developed by Wilson et al. (2020), where those authors concluded that smaller prey with higher agility would force larger predators to run along curved paths that do not allow them to use their superior speeds, and therefore could be a critical parameter for escaping predation. Moreover, in their study of the goshawk’s Accipiter gentilis (Linnaeus 1758) predation, Kane et al. (2015) showed a similar conclusion: the prey’s sharp sideways turns caused the goshawk to lose visual fixation on the prey and thus decreased their predation success.
The increasing number of hornets present in front of the hive strongly affected both hornet behaviour and honey bee behaviour on entering and leaving their hive. First, an increased number of hornets reduced the speed of bees leaving the hive, suggesting more hesitancy from bees going out to forage. The number of hornets also enhanced the speed and the trajectory curvature of honey bees entering the hive, so the bees are ‘racing’ into the hive, and choosing an unpredictable flight path – both of which may reduce their chances of capture. Very interestingly, this result fits with the adaptive behaviour of Apis ceranaforagers to escape V. velutina predation (Tan et al., 2007).Apis cerana are native honey bees from Asia an coexist withV. velutina in this region where both the Asian honey bee and Asian hornet are native. The similar behaviour of bees from Europe (our study) and Asia in the presence of an abundance of hornet predators suggests that increasing flight speed of honey bees entering the hive, i.e. the preferential prey of hornets in comparison with honey bees leaving their hive, would improve bee survival and limit hornet predation success. Second, increasing the number of hornets reduced the variance in flight patterns for bees and hornets. This could be advantageous to hornets at first as they have a lower range of bee flight trajectories to tackle and anticipate for predation success.
Although based on one single beehive, this study illustrates the interest of those observation and analysis methods, as they provided unique and very useful data, that allowed the observer to accurately witness complex phenomena congruent with the literature, and provided interesting leads for further studies. The automatic processing method providing ”Filtered video sequences of potential predation” from ”RGB-D sequences” represents a very useful tool for video-based data collection in ecology. Some improvement points can be recommended for future studies. External uncontrolled events (e.g. other flying insects coming into the field of view of the video camera, extreme weather episodes, overcrowded flight area) are likely to induce biases in the final statistics for the following reasons: target miss detections, failure during tracking, erroneous built-in depth estimation by the stereo-camera. Therefore, relative analysis and examination of trends is more conservative. For absolute figures, for example of ”hornet predation success rate”, the data would benefit from confirmation with complementary studies. This method could also allow to go deeper in the description of such behaviours: for example, through marking hornets’ hunters of different ages (different hunting experience) and colonies, it could be possible to describe individual learning, and to detect behavioural colonial differences. The same way through marking escaped bees, their learning capacity could be further evaluated.
In summary, 3D flight analysis of both predator and prey has demonstrated the characteristics of the bees’ flight that change (speed and trajectory curvature) as predatory pressure at the hive increases, suggesting avoidance behaviour. It has also shown that curvature of the bees’ flight has more effect than flight speed on hornet predatory success. As honey bee colonies in Europe are now under considerable pressure from this predator, it would be of great interest for future work to focus on whether different bee colonies show different flight behaviours making them more or less resilient to attack by this voracious predator.