Introduction
Crop farming can be challenging within the range of African elephants Loxodonta spp., and ongoing land-use change exacerbates encroachment of agriculture into elephant habitats (Mmbaga et al., Reference Mmbaga, Munishi and Treydte2017; Puyravaud et al., Reference Puyravaud, Gubbi, Poornesha and Davidar2019). Elephants enter farmlands and feed on crops mostly at night (Gunn et al., Reference Gunn, Hawkins, Barnes, Mofulu, Rachel and Norton2013; Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016), often leading to negative human–elephant interactions. Several strategies have been developed to promote coexistence, including biological methods (Vollrath & Douglas-Hamilton, Reference Vollrath and Douglas-Hamilton2002; Nelson et al., Reference Nelson, Bidwell and Sillero-Zubiri2003; King, Reference King2010; King et al., Reference King, Lala, Nzumu, Mwambingu and Douglas-Hamilton2017). However, often these strategies are only effective temporarily or do not meet people's expectations in terms of their ability to prevent crop damage by elephants (Nelson et al., Reference Nelson, Bidwell and Sillero-Zubiri2003; King et al., Reference King, Lala, Nzumu, Mwambingu and Douglas-Hamilton2017; Dror et al., Reference Dror, Harich, Duangphakdee, Savini, Pogány and Roberts2020).
Honeybees Apis mellifera are increasingly being used to protect crops from elephants (Vollrath & Douglas-Hamilton, Reference Vollrath and Douglas-Hamilton2002; Soltis et al., Reference Soltis, King, Douglas-Hamilton, Vollrath and Savage2014; Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016; Cook et al., Reference Cook, Parrini, King, Witowski and Henley2017; King et al., Reference King, Lala, Nzumu, Mwambingu and Douglas-Hamilton2017). Apis mellifera adansonii in West and Central Africa and Apis mellifera scutellata in East and Southern Africa (Fletcher, Reference Fletcher1978; Engel, Reference Engel1999) have a reputation of particularly aggressive behaviour, and their stings can kill animals (e.g. humans: Fletcher, Reference Fletcher1978; Soumana et al., Reference Soumana, Kamaye, Mamane, Mamoudou, Samailla and Moussa2016; waterbuck Kobus ellipsiprymnus: Barnes et al., Reference Barnes, Diego and Danquah2005; goats Capra spp.: Karidozo & Osborn, Reference Karidozo and Osborn2005). They repel intruders crossing their defensive perimeters (Lecomte, Reference Lecomte1961) by spreading pheromones (Wright et al., Reference Wright, Spencer, Cook, Henley, North and Mafra-Neto2018), buzzing or stinging (Soltis et al., Reference Soltis, King, Douglas-Hamilton, Vollrath and Savage2014; King & Raja, Reference King and Raja2016; King et al., Reference King, Pardo, Weerathunga, Kumara, Jayasena, Soltis and de Silva2018). The effect of pheromone release on savannah elephants Loxodonta africana has been demonstrated in Greater Kruger National Park, South Africa (Wright et al., Reference Wright, Spencer, Cook, Henley, North and Mafra-Neto2018). In Kenya, farms protected by beehive fences were more productive than unprotected farms as elephants succeeded only in 20% of their attempts at breaking such fences (King et al., Reference King, Lala, Nzumu, Mwambingu and Douglas-Hamilton2017). Similarly, in Gabon empty hives and hives with low bee activity (< 40 bee movements per minute; a bee movement being defined as a bee exiting or entering the hive) did not deter elephants, whereas active hives (40–60 bee movements per minute) did (Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016).
Bees are predominantly diurnal insects and only a few species fly at night (Theobald et al., Reference Theobald, Greiner, Wcislo and Warrant2006). For example, A. m. adansonii can take advantage of moonlight to forage at night (Fletcher, Reference Fletcher1978; Theobald et al., Reference Theobald, Greiner, Wcislo and Warrant2006), and when disturbed, A. m. scutellata have been observed to swarm from beehives to repel elephants during the night as well as during the day (King, Reference King2013). Farmers are often reluctant to adopt honeybees as elephant deterrents (King, Reference King2010; Noga et al., Reference Noga, Kolawole, Thakadu and Masunga2015; King et al., Reference King, Lala, Nzumu, Mwambingu and Douglas-Hamilton2017), and in Thailand it was reported that A. mellifera and Apis cerana were not aggressive towards Asian elephants Elephas maximus when disturbed during the day or at night (Dror et al., Reference Dror, Harich, Duangphakdee, Savini, Pogány and Roberts2020). These geographical and temporal variations in the behaviour of bees call for site-specific research to validate the efficacy of honeybees as potential elephant deterrents. This should be done before investment in beehive fences is promoted.
Encroachment of agricultural areas into elephant habitat around Campo–Ma'an National Park (Cameroon) has intensified in recent years, increasing competition between people and elephants over space and resources (MINFOF, 2014). We experimentally assessed the aggressiveness of disturbed A. m. adansonii at different times of day to determine whether they could be used to deter intruding elephants. In the first study of this kind in this area, we artificially disturbed and recorded the behavioural responses of A. m. adansonii during daytime and night-time periods to assess their potential efficacy for use in beehive fences to protect crops. We evaluated three indicators of honeybee efficacy in protecting crops from simulated elephant visits: (1) the activity level of colonies (measured as the frequency of bee movements at the hive entrance), (2) the level of aggressive behaviour of the colonies (measured as the mean distance from hives at which honeybees showed defensive behaviour), and (3) the bees’ response in the form of a defensive flight when disturbed by an intruder.
Study area
We conducted our field experiments in Mabiogo (Fig. 1), one of 162 villages in the Campo–Ma'an conservation area in southern Cameroon, which includes Campo–Ma'an National Park (264,064 ha). Approximately 111,000 people of various socio-cultural backgrounds live in the conservation area, all of which rely on agriculture and forest products, including wild honey, for their livelihoods (Tiani et al., Reference Tiani, Akwah, Nguiébouri and Colfer2005; MINFOF, 2014). Staple food crops are grown during the two annual rainy seasons, and farmers experience interactions with wildlife from the Park, including an estimated population of 544 (range: 425–695) free-ranging forest elephants Loxodonta cyclotis (MINFOF, 2014). The Park is unfenced and beekeeping is unusual in the area. However, interactions between elephants and wild honeybee colonies are expected to occur in the forest. The mean annual precipitation is c. 2,500 mm, the mean annual temperature is 22–28 °C, and the area maintains high humidity throughout the year. Many rivers and swamps are present in the area and the vegetation consists of trees and herbaceous flowering plants (Tchouto, Reference Tchouto2004).
Fig. 1 Location of Mabiogo, the study site in the Campo–Ma'an conservation area, Cameroon.
Methods
Data collection
We collected data during 24 June–10 August 2019. In 2017, we had constructed a total of 22 Kenyan top bar hives (Supplementary Plate 1) following a previous conceptual model (King, Reference King2014), and had distributed these to two farmers to start apiculture. We numbered the hives H1–H22 and placed them at the edges of the farms. We set the distance between neighbouring hives at 10 m (King, Reference King2014). Two years after we set up the hives, only six hives, colonized at different time periods, had active colonies (H1, H6, H12 and H14 from one farmer H8 and H17 from the other farmer) and we treated each colony as an experimental unit. For safety reasons we wore beekeeper suits, gloves and rubber boots when assessing bee activity (Nouvian et al., Reference Nouvian, Reinhard and Giurfa2016). At each farm we collected data regarding both visual and physical disturbances at different times during the day (morning: 05.00–12.00, noon: 12.00–14.00, afternoon: 14.00–18.00) and at night (evening: 18.00–21.00, night: 21.00–00.00).
Activity level of the colonies
To assess whether the activity level of the colonies (a measure of defensive behaviour) would affect their ability to deter elephants, we recorded 5-minute videos of bees entering and leaving each beehive (Woyke, Reference Woyke1992; Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016) using a high-resolution infrared camera (Sony HDR-SR12, Sony, Tokyo, Japan) that enabled us to record at night, for a total of six recordings per hive. We only included videos from which we were able to obtain counts of bees. We calculated the activity level using the following formula (Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016):
$${\rm Number\ of\ bee\ movements}/{\rm minute\ } = ( {{\rm Numbe}{\rm r}_{{leaving}}_{} + {\rm Numbe}{\rm r}_{{entering}}} ) /5$$Defensive reaction of honeybees to an approaching observer
Hives are guarded by soldier bees who control the flow of bees in and out of the hive, ward off impending threats and alert the colony in the event of approaching threats (Breed et al., Reference Breed, Guzm and Hunt2004; Nouvian et al., Reference Nouvian, Reinhard and Giurfa2016). To assess the ability of A. m. adansonii to repel encroaching intruders (using vision or scent), we walked at a constant pace from random positions towards the hive entrance and stopped when an attack occurred. We measured the distance between the hive and the position of the observer to determine the defensive perimeter of the hives. We considered an attack to be the circular movement of bees around the person approaching the hive. Bee movements were passive (inoffensive) or active, potentially resulting in a bee sting (Lecomte, Reference Lecomte1961; Nouvian et al., Reference Nouvian, Reinhard and Giurfa2016).
Response of honeybees to a physical threat
Physical disturbance triggers the defensive behaviour of honeybees (Fletcher, Reference Fletcher1978; Breed et al., Reference Breed, Guzm and Hunt2004; King, Reference King2010). When elephants walk through a beehive fence they cause multiple hives to swing, leading to the bees releasing an alert pheromone, flying out or targeting and repelling intruders (King et al., Reference King, Douglas-Hamilton and Vollrath2007; King, Reference King2010). To assess the bees' defensive response to a simulated disturbance, we used a stick to mimic an elephant entering the farm and noted whether at least one bee flew out of the hive beyond a distance of 1 m. We coded the responses in a binary fashion according to whether bees flew > 1 m away from their hive or not (i.e. flying ≤ 1 m from their hive).
We waved the stick near the entrance of the hive for 1 minute, and then gently touched the guard bees sitting at the entrance of the hive, without introducing the stick into the hive. We noted the start time of each disturbance to account for the effects of weather parameters on the bees' activity (Lecomte, Reference Lecomte1961; Breed et al., Reference Breed, Guzm and Hunt2004). To control for the possible influence of temperature on bee activity, we measured the ambient temperature and that within the hives, the latter using a thermal probe placed inside the hive prior to the physical disturbance of the colony and removed after data collection (Burrill & Dietz, Reference Burrill and Dietz1981). We used a mini weather station to record air humidity.
We performed two successive disturbances at every visit for each hive, separated by a 5-minute break. Each sequence of data collection at a hive lasted c. 13 minutes, therefore totalling c. 31 minutes per hive per visit. At the end of the second sequence we recorded the internal temperature of the hive before extracting the probe. During the disturbance, a field assistant recorded the time, air humidity and whether or not bees flew from the hive, whilst remaining at least 4 m away from the hive, which corresponds to the minimum distance between hives when constructing beehive fences (King, Reference King2010). To allow the colonies to calm down during the 5-minute break, we moved 10 m away from the hives. When bees from the disturbed colony remained agitated beyond the 5-minute break, we chased them away using a smoker before initiating the next sequence of data collection. We took precautions to avoid modifying the behaviour of the colonies with smoke (Woyke, Reference Woyke1992).
Data analysis
We used repeated ANOVAs to assess the differences in activity level between colonies and between times of day, followed by a Tukey honestly significant difference test for post hoc analysis to compare colonies. In the analysis of the defensive perimeter before the physical disturbance we considered all values equal to 0 m to be the dormant state of the hives and omitted them from the analysis to avoid minimizing the mean defensive zone of the colonies, which could be misinterpreted by farmers. To assess the temporal variation of the defensive perimeter, we used the linear mixed-effect function of the lme4 package in R 3.6.3 (Bates et al., Reference Bates, Maechler, Bolker and Walker2015; R Core Team, 2020) fitted by restricted maximum likelihood. We considered the response variable to be the distance at which the defensive reaction was observed, and included the day (i.e. date) of the observation as a random term because we took repeated measures on the same day. For the explanatory variables we used the time of day (categorical variable with five values: morning, noon, afternoon, evening and night), the colonies (categorical variable with six values, representing the individual hives) and the order of the test (first or second approach). When we found a significant effect of time of day or colony, we performed a Tukey honestly significant difference test to compare the mean distance at which defensive behaviour was observed between different times of day and between colonies.
We used χ 2 tests to assess the dependency between disturbances and the occurrence of honeybees flying > 1 m away from the hive. We used the lme4 package in R (Bates et al., Reference Bates, Maechler, Bolker and Walker2015) fitted by maximum likelihood (Laplace approximation) with a binomial distribution and a logit link to assess the effects of the colony, temperature in the hives (a continuous variable) and time of day on whether or not bees flew > 1 m away from the hive. We used likelihood ratio tests to assess the significance of the effects of time of day and colony. We included the day (date) as a random term because we took repeated measures on the same day, and considered colony, time of day and temperature inside the hive to be fixed variables in the model. We used Tukey honestly significant difference tests when we observed differences between different times of day or between colonies. We performed statistical analyses with the significance level set at 0.05.
Results
Activity level of the colonies
The activity of bees prior to physical disturbance differed significantly between diurnal and nocturnal periods (F (1,154) = 565, P < 0.001) and between colonies (F (5,72) = 7.45, P < 0.001). Colonies were active during the day with a mean of 49 bee movements per minute (range: 35.69 ± SD 11–69.55 ±SD 16.53) and were inactive during the night. Colony H14 was significantly more active than colonies H1, H8, H12 and H17 (Tukey honestly significant difference test at 95% CI, all adjusted P < 0.05), and colony H6 was significantly more active than colony H17 (Tukey honestly significant difference test at 95% CI, adjusted P = 0.043). All other pairs were not significantly different in terms of bee activity (Tukey honestly significant difference test at 95% CI, all adjusted P > 0.05; Fig. 2).
Fig. 2 Distribution of the number of African honeybee Apis mellifera adansonii movements per minute (entering and exiting) at the entrance of each hive during daytime periods. The horizontal lines within the boxes represent the median, and the boxes the interquartile range (between the 25 and 75 percentiles). The upper and lower whiskers represent the maximum and minimum values, respectively, that are within 1.5 times the interquartile range. The circles denote outliers.
Defensive reaction of honeybees to an approaching observer
We performed 276 approaches, of which 20% yielded a defensive response and 80% did not. Of the 134 nocturnal approaches (48.5% of all approaches), 95% yielded no reaction, and we recorded a defensive response rate of 5% in the evening. Of the 142 diurnal approaches (51.5% of all approaches), 65.5% yielded no reaction and 34.5% yielded a reaction. The mean defensive perimeter across different times of day and colonies was 4.05 ± SD 2.5 m. The distance from which bees responded to an approaching observer differed significantly between times of day (F (4,25) = 4.716, P = 0.006; Table 1), with a larger defensive perimeter in the evening than at other times (Table 1). The mean defensive perimeter also differed significantly between colonies (F (5,25) = 8.692, P < 0.001; Table 1), with colony H1 having the largest defensive perimeter (7.0 ± SD 2.8 m). There was no difference in defensive perimeter (F (1,54) = 1.55, P = 0.219) between the first (mean 3.58 ± SD 2.58 m) and second approach (4.40 ±SD 2.34 m).
Table 1 Mean defensive perimeters (m) of African honeybee Apis mellifera adansonii colonies at different times of day in response to an approaching observer, and corresponding Tukey honestly significant difference (HSD) tests. All tests performed at P = 0.05, and a, b, c and d refer to the result of the test between colonies: hives with the same letters were not significantly different from each other.
Response of honeybees to a physical threat
Of the 276 disturbances, 51% (n = 142) were diurnal, and 49% (n = 134) were nocturnal. The majority (63.4%, n = 175) of the disturbances resulted in a defensive flight of honeybees (χ 2 = 19.841, df = 1, P < 0.001), with 67% (n = 117) occurring during the day and 33% (n = 58) at night. When assessing whether bees responded to a threat, we found a significant difference between times of day (χ 2 = 20.2, df = 1, P < 0.001; Fig. 3a) and colonies (χ 2 = 120, df = 1, P < 0.001; Fig. 3b). The results from our mixed model showed that on average, compared to at night, bees flew more during the morning (3.842 ± SE 0.872, P < 0.001), noon (4.732 ± SE 1.039, P < 0.001) and afternoon (4.279 ±SE 1.053, P < 0.001). Response to a threat did not differ between morning, noon and afternoon (all pairwise comparisons P > 0.05). Similarly, compared to colony H1, bees from colonies H17 (–4.897 ± SE 0.930, P < 0.001) and colony H8 (–2.708 ± SE 0.783, P = 0.005) responded less frequently with a defensive flight when threatened. Bees from colony H12 showed a defensive flight more often than those from colonies H17 (5.547 ± SE 1.115, P < 0.001) and H8 (3.359 ± SE 1.069, P = 0.015), and bees from colony H6 responded more than those from colonies H8 (2.642 ± SE 0.787, P = 0.007) and H17 (4.830 ± SE 0.956, P < 0.001). All other pairwise comparisons were non-significant (all P > 0.05).
Fig. 3 Per cent response (defined as flying > 1 m away from the hive) of African honeybees to physical disturbances (a) for different colonies and (b) for different times of day.
Discussion
We found that honeybee colonies differed in their activity level and defensive behaviour when disturbed. In addition, honeybee colonies were only active during the day and their defensive perimeters were greater in the morning and evening when the bees appeared to be more sensitive to disturbance. These findings suggest that beehive fences may be less effective at deterring intruders at night.
Activity level of the colonies
We assessed the activity level of A. m. adansonii as an indicator of aggressive behaviour and found that the activity levels of most colonies were above the requirements for use as beehive fences. Four colonies exhibited daytime activity of 40–60 bee movements per minute, levels that have been found to be effective for deterring forest elephants in Gabon (Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016). However, activity levels of two colonies were below the required range for an effective deterrent. This suggests that when setting up beehive fences, colonies should be selected for inclusion based on their activity levels (Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016).
At night all colonies were clustered at the entrances of their hives and visibly inactive because of decreasing temperature (Burrill & Dietz, Reference Burrill and Dietz1981) and increasing humidity (Supplementary Fig. 1). We observed no bees flying prior to us disturbing the colonies at night. This corroborates findings from a study in Thailand using A. m. scutellata and A. cerana (Dror et al., Reference Dror, Harich, Duangphakdee, Savini, Pogány and Roberts2020). However, it contradicts observations of A. m. adansonii foraging at dusk, under low light intensity. Had the bees been more active at night, it would have increased their potential use in beehive fences, as most elephant intrusions into agricultural areas occur at night (King, Reference King2010; Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016).
Defensive reaction of honeybees to an approaching intruder
In response to an approaching intruder, bees were mostly inactive, except in the morning and twilight periods when they were more likely to fly and attack intruders. Similar patterns of aggressive behaviour in A. mellifera have been reported previously (Lecomte, Reference Lecomte1961; Woyke, Reference Woyke1992). This finding is not surprising because most foraging bees exit the hive in the morning and return in the evening (King, Reference King2010). Hives are guarded by mature foragers who are more experienced and produce more pheromones than younger individuals (Nouvian et al., Reference Nouvian, Reinhard and Giurfa2016). We argue that beehive fences would be more effective during the morning and dusk than during other times of day because mature foragers help defend the hives during these periods. These two periods have also been reported as the times when elephants frequently enter or leave plantations (King, Reference King2009; Gunn et al., Reference Gunn, Hawkins, Barnes, Mofulu, Rachel and Norton2013; Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016). Active colonies could thus potentially repel elephants approaching during the evening because, if disturbed, the bees would probably fly out and attack the elephants (it is not completely dark until 19.00 in this area).
Response of honeybees to a physical threat
Our results showed that disturbed bees were more likely to fly out from the hive and repel intruders during the day than at night. All colonies reacted vigorously to physical disturbance during the daytime, with bees flying in all directions to identify and sting the intruder. Similar responses of bees to physical threats during the daytime and at twilight have been reported previously (Woyke, Reference Woyke1992; Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016; King et al., Reference King, Lala, Nzumu, Mwambingu and Douglas-Hamilton2017). However, their decreased level of defensive behaviour after dusk reduces their effectiveness in repelling animals with a high cognitive capacity such as elephants (Dror et al., Reference Dror, Harich, Duangphakdee, Savini, Pogány and Roberts2020). At night, physical disturbances resulted in bees falling to the ground because they were unable to fly; they had to walk towards the support of the hive to climb up and return to it. The bees buzzed loudly in response to such night-time disturbances, except during periods of bright moonlight, when no buzzing occurred. Although forest elephants in Gabon avoided colonies with high levels of activity (Ngama et al., Reference Ngama, Korte, Bindelle, Vermeulen and Poulsen2016), our results suggest that the inactivity of bees at night could be noticed and exploited by forest elephants through breaches in the fences at night, particularly if the elephants are exposed repeatedly to such bee behaviour (Dror et al., Reference Dror, Harich, Duangphakdee, Savini, Pogány and Roberts2020).
Towards the end of the study we noted that bees from the smallest colonies (H8, H12 and H17) flew inside their hives when disturbed rather than away from the hive and towards the source of disturbance, even during the daytime. This was unexpected as bees are usually aggressive during the daytime. We argue that repeated disturbances could reduce the aggressiveness of colonies, especially the smallest ones, because of the loss of mature guards, leaving the hives inadequately protected by less experienced guards (Nouvian et al., Reference Nouvian, Reinhard and Giurfa2016). In contrast, larger colonies such as H1, H6 and H14 (Supplementary Plate 2b) were more reactive and never inactive during the day. Colony size thus affects the bees’ response to a threat, although hives can become less reactive especially when the queen bee is not in the hive (Lecomte, Reference Lecomte1961; Woyke, Reference Woyke1992; Supplementary Plate 2a).
In summary, our findings highlight the need to combine other mitigation methods with beehive fences to improve their effectiveness (Nelson et al., Reference Nelson, Bidwell and Sillero-Zubiri2003; King, Reference King2010). Although we observed low levels of defensive flights around noon and at night, bees may still be able to deter elephants at these times as buzzing and pheromones could be part of their defensive mechanism and may have a deterrent effect on elephants. More research into these aspects of bee behaviour is required to improve our understanding of the efficacy of beehive fences.
The predictive capacity of our study is limited because our experimental design did not involve beehives being disturbed by actual elephants. Nevertheless, it provides valuable insights into the potential effectiveness and limitations of beehive fences to deter forest elephants and reduce crop losses that affect people living near protected areas. Our findings on the threat response of A. m. adansonii thus have the potential to facilitate informed decision-making regarding the use of beehive fences to address crop damage by elephants.
Acknowledgements
We thank the Zoo de Granby, MITACS Accelerate, Quebec Centre for Biodiversity Sciences and Concordia University for financial support; Valerie Michel of Zoo de Granby for helping with data collection and data entry; Alexander Boucher and Alys Granados for editing the manuscript; and the Ministry of Forestry and Wildlife of Cameroon, particularly Benjamin Sock, former Conservator of the Campo–Ma'an National Park, for enabling the work to take place.
Author contributions
Conception and design: IBD, RBW; data collection: IBD; data analysis: IBD, RBW; writing: IBD; revision: all authors.
Conflicts of interest
None.
Ethical standards
Data collection and safe handling of animals were conducted in accordance with the Animal Research Ethics certificate provided by Concordia University (Protocol number 30003983). The research protocol was reviewed and approved by the Ministry of Forestry and Wildlife of Cameroon prior to carrying out the experiment. The research otherwise abided by the Oryx guidelines on ethical standards.
Open access
