The European honey buzzard ( Pernis apivorus ) as an ally for the control of the invasive yellow‐legged hornet ( Vespa velutina nigrithorax )

Abstract

BACKGROUND

Biological control in integrated pest management (IPM) often overlooked avian predators until the emergence of the ecosystem services approach. Birds are now recognized as key regulators of pest populations in agroforestry landscapes due to their high mobility. The invasive yellow‐legged hornet, introduced into Europe in 2004, threatens agriculture, beekeeping and native pollinators. We aimed to determine whether European honey buzzard attacks on yellow‐legged hornet nests reduce the densities of individuals (workers) in summer and full‐grown colonies in November around the raptor's nests in southwestern Europe.

RESULTS

We analyzed honey‐buzzard foraging patterns of 11 individuals during breeding using trail cameras and GPS emitters to locate attacked hornet nests. The average mode distance from raptor nests to the attacked hornet nests was 1234.7 m, with 89.3% of attacked nests destroyed. We assessed the change in the abundance of hornet workers and in the density of full‐grown nests over distance in the vicinity of 17 honey‐buzzard nests and 10 control points, finding a significant decline of −0.000116 workers h⁻¹ m⁻¹ within 1000 m of a raptor nest. This impact intensified as the breeding season progressed. However, no significant effect on the density of full‐grown hornet nests was observed.

CONCLUSION

These results are of interest for the management of the exotic hornet, at least on the abundance of workers and at a small scale in the proximity of honey‐buzzard nests. These raptors should be considered allies in the fight against hornet populations and included in IPM programmes as a native controller of the pest. © 2025 The Author(s). Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

The predatory effect of the honey‐buzzard affects the reproductive performance of Asian‐hornet colonies, decreasing the density of workers over distance and time. The foraging distances of the honey‐buzzard concentrates within the first 2000 m from nest, which supports the results observed. Active goshawk breeding territories limit the honey‐buzzard predatory effect.

1. INTRODUCTION

Since its emergence in the late 20th Century, integrated pest management (IPM) has included, as a central component, the biological control of pests by means of predators, pathogens and/or parasitoids, to reduce pest population densities. ¹ , ² For a long time, biological control in IPM neglected avian predators as natural enemies until the emergence of the ecosystem services approach, which reshaped the paradigm of IPM practices. ³ , ⁴ IPM is often defined as a holistic strategy to manage pests and maintain their population densities within limits compatible with production and with minimal pesticide application. ⁵ , ⁶ However, humans continue to use pesticides each year to reduce pest population densities. ⁷ , ⁸ Therefore, greater effort must be made to redirect pest management policies towards a biodiversity‐friendly approach, which enhances pest regulation ecosystem services while improving biodiversity and sustainability of our production systems. ⁹

Avian predators are valuable allies in the IPM of crop pests. The ability to fly confers them high mobility, which results in an increased connectivity between patches that might otherwise be isolated. ¹⁰ It has been proven that birds are important regulators of pest populations in agroforestry heterogeneous landscapes. ¹¹ Most pest‐regulating bird species consume arthropods, either as specialists or opportunistic predators. ⁴ , ¹² , ¹³ Nyffeler et al. ¹⁴ estimated an annual consumption of 400–500 million tonnes of arthropods by birds globally, and ≈7% of that consumption occurring in agroecosystems. García et al. ¹⁵ using bird exclusion experiments, proved that the absence of avian predators resulted in increased crop damage by aphids. Despite some experimental and observational studies, our current understanding of the ability of raptor species to regulate agricultural insect pests remains limited. ¹¹

In this study we assess the influence of the European honey buzzard (Pernis apivorus L. 1758) (hereafter, honey‐buzzard) on the abundance of the yellow‐legged hornet (Vespa velutina Lepeletier 1836, var. nigrithorax du Buysson 1905) in agroforestry ecosystems of southwestern Europe. The yellow‐legged hornet is an exotic invasive species accidentally introduced into Europe in 2004. ¹⁶ , ¹⁷ Its presence causes losses to agriculture and beekeeping, and poses a threat to the native pollinators of the ecosystems colonized. ¹⁸ , ¹⁹ Adult hornets feed on carbohydrate sources such as sap and fruits, damaging fruit crops and decreasing production in vineyards. ²⁰ , ²¹ , ²² Larvae require high protein sources to grow, so adult hornets scavenge and hunt insects to feed their brood. ²³ Villemant et al. ²⁴ assessed that ≈60% of insect prey were honey bees. Entire hives can be taken down in weeks, resulting in the death of the whole colony and significant losses to the beekeeping sector. ¹⁹ In 2014, the European Union included the yellow‐legged hornet on its list of invasive exotic species (UE1143/2014).

The honey‐buzzard is the only raptor specialized in consuming colonies of eusocial vespids in Europe. ²⁵ It has rapidly incorporated the yellow‐legged hornet in its diet, becoming the second most‐consumed prey. ²⁵ , ²⁶ , ²⁷ The honey‐buzzard locates underground colonies of vespids and digs out the combs to feed on the brood (larvae and pupae), or carries the combs to the nest to feed its nestlings. ²⁸ , ²⁹ The larvae play a crucial role as a digestive caste in eusocial vespids, ³⁰ , ³¹ , ³² and therefore, the predation of the brood by honey‐buzzards is likely to lead to the collapse of the entire colony.

This predator–prey relationship differs from other systems of biological control of crop pests by birds. Most studies ¹¹ , ¹³ , ¹⁵ focus on non‐eusocial prey insect species, which typically have several generations per year and can re‐colonize crop fields after the predation pressure by birds. Eusocial vespids show annual cycles in which the only reproductive individuals are males and the queen, and the colonization of new areas occurs only by new queens after hibernation. Therefore, the elimination of a colony results in the disappearance of vespids from the area until the following year, when other gynes colonize the area. ²³ , ³³ , ³⁴ This makes honey‐buzzards optimal candidates to be studied as biological controllers of yellow‐legged hornets.

The honey‐buzzard is a migratory forest‐dwelling raptor that spends a short period in Europe, mainly from early May to late August, to breed. Asian hornet gynes emerge from hibernation in early spring (March or April), present juvenile nests in summer (July and August), and develop reproductively capable nests (full‐grown nests) by autumn. The aim of this study is to determine whether honey‐buzzard attacks on yellow‐legged hornet nests reduce the abundance of individuals (workers) in summer (July and August) and the density of full‐grown mature colonies in autumn (November) around the raptor's nests. We address three specific objectives and pose hypotheses (H) and predictions (P). (i) Estimate the density of yellow‐legged hornet nests attacked and the extent of destruction inflicted by honey‐buzzards around the raptor nests. As a central place forager during breeding season, we hypothesize (H1) that the foraging activity would be focused on the vicinity of the honey‐buzzard nest because the raptor needs to deliver the hornet combs to its nestlings. We expect (P1.1) a higher density of attacked hornet nests close to the raptor nest and a rapid decrease in the density of attacked hornet nests as distance from the raptor nest increases. Additionally, we expect (P1.2) the maximum distance of attacks to increase as the breeding season progresses, as a consequence of the depletion of available hornet nests near the raptor's nest. (ii) Study the changes in abundance of yellow‐legged hornet workers in the honey‐buzzard breeding territory compared to those areas excluded from the raptor attacks. The Eurasian goshawk (Accipiter gentilis L. 1758) (hereafter goshawk) is the main predator of honey‐buzzard adults and nestlings in Europe, and honey‐buzzards avoid proximity to the active goshawk nests. ³⁵ We considered active goshawk breeding territories as experimental exclusion areas for honey‐buzzards. We hypothesize (H2) that the elimination of yellow‐legged hornet nests by honey‐buzzards is more intense in the vicinity of active honey‐buzzard nest than in the vicinity of active goshawk nests, in which the honey‐buzzard would reduce its hunting activity. This effect would increase as the raptor's breeding season progresses as the number of colonies eliminated increases, accumulating the predatory effect as it depletes the number of colonies in the area. Therefore, we expect (P2.1) that the abundances of workers in the vicinity of honey‐buzzard nests are lower than those surveyed in the vicinity of goshawk nests. We also expect (P2.2) the rate of capture of workers to decrease gradually in honey‐buzzard areas throughout the breeding season until late August, but not in goshawk areas. (iii) Assess the effect of honey‐buzzards on the density of full‐grown mature yellow‐legged hornet nests in autumn, when colonies can produce reproductive individuals (males and gynes). We hypothesize (H3) that the honey‐buzzard attacks on hornet nests during summer reduce the density of nests that reach the end of the reproductive cycle and produce reproductive individuals (full‐grown mature nests) in autumn. We expect (P3) to find lower densities of full‐grown yellow‐legged hornet nests closer to honey‐buzzard nests and higher densities closer to goshawk nests. Thus, we expect the rate of capture of yellow‐legged hornet workers and the density of full‐grown nests to increase with distance from honey‐buzzard nests and to decrease with distance from goshawk nests. In light of the results, we discuss the implications of the predation by honey‐buzzards on the management of the invasive hornet.

2. MATERIAL AND METHODS

2.1. Study area and species

The study area is located in the north‐western Spain (42° 20′ N, 8° 47′ W), with an extension of 670 km² (Supporting information, Fig. S1). Climate is mild humid‐oceanic, with annual average precipitation of 1402 mm and annual average temperature of 14.2 °C. ³⁶ The landscape is hilly and presents an average altitude of 213 m above sea level (a.s.l) (range 0–646 m a.s.l). It conforms a heterogenous mosaic of forest patches (50.9%) in the middle‐upper part of the slopes, and farms with crops and scattered houses (35.5%) in the valley bottoms. ³⁷ The remaining 13.2% corresponds to urbanized areas (cities, towns and villages). These forest patches are dominated by non‐native eucalyptus plantations (Eucalyptus globulus Labill 1800) but there are mixed patches with native species such as the pedunculate oak (Quercus robur L. 1758) or the maritime pine (Pinus pinaster Ait. 1789).

The European honey‐buzzard is a forest‐dwelling migratory raptor distributed across most of Europe during the breeding season and spends the nonbreeding season in the sub‐Saharan wintering areas. ³⁵ The breeding season starts in early May, right after the pre‐nuptial migration, and ends in late August. The Iberian Peninsula hosts low‐density breeding populations located mainly in the forested areas of the northern Spain and Portugal. ³⁸ The first eggs – usually two – are laid in the second half of May, on alternating days. ²⁷ , ²⁸ Hatching occurs after 33–35 days of incubation. The chicks are fed by adults until they reach complete independence, around 65 days. ²⁸ , ²⁹

The honey‐buzzard is a secretive and elusive breeder whose active nests are difficult to locate. ²⁸ , ³⁹ Adults and nestlings of honey‐buzzards suffer predation from goshawks, so their nests tend to be located far away from active nests of this predator. ⁴⁰ For our experimental design, we assumed that active goshawk territories are exclusion areas or areas with less predatory activity of honey‐buzzards that can be used as experimental controls.

The yellow‐legged hornet is native to Southeast Asia but was introduced accidentally in Europe in 2004 as a consequence of the maritime transport of gardening supplies containing hibernating gynes. ¹⁶ , ¹⁷ The first colony at the study area was recorded in 2014; ⁴¹ 1 year later the Spanish government included the yellow‐legged hornet in the Catalogue of Invasive Exotic Species ⁴² and approved a national pest management strategy specifically targeting the yellow‐legged hornet. ⁴³

The biological cycle of the yellow‐legged hornet is annual. ²³ Mature colonies release reproductive individuals (gynes and males) in autumn. After mating, the gynes disperse and search for shelter where they can hibernate, while their parental colonies die during winter. After hibernation, the gynes emerge and start building the embryo nests around March and April, frequently underground. The first workers can be detected around mid‐May. The primary nest grows during late spring and early summer and usually the colony requires more space, prompting a move to a new location in the open during the summer, where they build the secondary nest. This secondary nest will ultimately mature and grow enough to produce reproductive individuals for the next breeding season (full‐grown nests). ³³ , ³⁴ , ⁴⁴ , ⁴⁵ Given the difficulty to ascertain if an aerial nest is secondary (relocated) or not, we chose to name aerial nests in autumn as mature.

2.2. Data collection

2.2.1. Location of active raptor nests

The surveys for active honey‐buzzard nests are part of a long‐term ecological research focused on the diurnal forest‐dwelling raptor guild since 2004 and continuing to the present. ²⁷ , ⁴⁶ The search effort primarily included goshawk, common buzzard (Buteo buteo L. 1758), Eurasian sparrowhawk (Accipiter nisus L. 1758), and honey‐buzzard. For 3 years (period 2020–2022), forest patches were systematically surveyed on foot to locate active honey‐buzzard and goshawk nests (more details in Rebollo et al. ²⁷ and Martín‐Ávila et al. ²⁵ ). During each breeding season, the status of each raptor nest and the breeding success were monitored using a high‐resolution camera (GoPro HERO) that we momentarily raised to the raptor nests by means of a pole. Monitoring was carried out several times during the breeding period (approximately once every 2 weeks), spending 10–20 min each time to obtain several photographs of the inside of the nest that the camera sent to our tablet. In the present study, we only considered the nests with fledglings to carry out our surveys to ensure that the activity of adult honey‐buzzards in the nesting territories persisted throughout the entire breeding season of the raptor (until late August).

2.2.2. Tracing honey‐buzzard's attacks using GPS data and photographic record

We captured six breeders from five nests in 2021 (the couple from one nest, one male from another, and three females from three nests), and five breeders from four nests in 2022 (the couple from one nest, one female from another, and two males from two nests). We collected GPS data from 11 different breeders, with four of them providing data for both breeding seasons. During the wintering season between 2021 and 2022, we lost signal from two females in Africa. The rest of the tagged individuals returned to the study area, usually choosing different nests and territories for breeding.

We conducted the capture attempts of breeders when the youngest chick had reached 14 days of age. At this stage, chicks could maintain their body temperature without parental assistance during the captures of breeders. We estimated the chicks' age based on pictures taken with the GoPro camera. Adults were captured using ‘dho‐gaza’ nets deployed 10–20 m from the nest tree, allowing direct view to the nest and in an open area to ensure the bird could fly unhindered. Honey‐buzzards were attracted using several lures (a stuffed remote‐controlled eagle‐owl or a live eagle‐owl or a live goshawk) placed on the ground. A sound system playing goshawk or owl calls was hidden next to the lure.

GPS emitters (Ornitrac‐20 model; Ornitela, Vilnius, Lithuania) were attached to the birds' bodies using Teflon harnesses created in situ to match each bird's specific body size. The harnesses were secured by a single stitched point made from cotton strands. This design allows the harness to last approximately 4 years and to detach safely from the bird's body when the cotton strands deteriorate. The GPS emitters were equipped with a solar powered battery, an accelerometer and GSM/GPRS 3G antennae. This configuration enabled them to transmit information remotely, eliminating the need for recapture to collect data and allowing for remote setting configuration. The logging frequency was set at 1 log every 2–15 min, depending on the battery level. Sufficient sun exposure allowed the device to charge the battery, resulting in better data resolution.

We installed trail‐cameras (S378 4G; Spromise, Shenzen, China) in the GPS‐tagged honey‐buzzards' nests to record prey deliveries to the nest by adults in real time. The trail‐cameras were placed on the surrounding branches, at 1.5–2 m from the nest, and were connected to an external battery to ensure coverage throughout the entire breeding period. The cameras were programmed to take a single picture when triggered by the movement of nestlings or adults, with a minimum delay of 1 min between pictures. Taking advantage of the moment of the camera installation, we checked the age of the chicks by interpolating wing length and body mass using the tables provided by Bijlsma ⁴⁷ to obtain the most accurate estimation of the number of days passed since hatching. Spromise cameras sent pictures via email, which allowed us to review the images daily to identify the moments when yellow‐legged hornet combs were delivered to the nest by the adults. Subsequently, we cross‐referenced the GPS data collected by the emitters before the comb deliveries to predict the locations where the honey‐buzzards gathered these combs (Fig. S2). Within the subsequent 24–72 h, a team of two to six surveyors visited these locations in the field, conducting systematic on‐foot surveys in search of the attacked hornet nest. Additionally, the location (subterranean or aerial) was documented and the extent of destruction inflicted on the hornet nest was assessed, using the following three categories. Almost intact: the hornet nest was active (workers and brood alive), two or more combs with larvae remained. Partially destroyed: the hornet nest was active, one comb with larvae remained. Evident signs of excavation by honey‐buzzards. Destroyed: the hornet nest was not active or the remaining workers staying in the burrow. No combs with brood remained.

In order to study the predatory pressure of the honey‐buzzard as a central place forager (specific objective 1), we studied the number of attacks on hornet nests around the honey‐buzzard nests. We segmented the area within 10 km radius around the honey‐buzzard nest in 500‐m‐wide circular bands, counted the number of hornet nests attacked within each band, and estimated the density of attacked hornet nests per unit area as we moved away from the honey‐buzzard nest (no of hornet nests attacked/area of each band).

2.2.3. Assessing the abundance of yellow‐legged hornet workers

During 2020, 2021 and 2022, we installed baited traps (Avispa'clac; Protecta) to estimate the abundance of yellow‐legged hornet workers in the raptor nesting territories. In 2020, we surveyed six honey‐buzzard territories; in 2021, four honey‐buzzard territories and four goshawk territories; and in 2022, five honey‐buzzard territories and four goshawk territories (Fig. S1). Throughout all years, we chose different raptor territories to avoid pseudoreplication between years. This decision does not allow studying the cumulative effects of honey‐buzzards over the years, but it does facilitate statistical analyses when, as in this case, the number of observations is low and honey‐buzzards do not usually repeat nest and breeding territory in subsequent breeding seasons. Each territory was surveyed by installing seven baited traps homogenously distributed as follows: one trap at 0 m from the raptor nest tree, three traps at 500 m from the nest tree and three traps at 1000 m from the nest tree (see Fig. S1). This design takes into account the flying distances of yellow‐legged hornets observed by Kennedy et al. ⁴⁸ in their study of radiotelemetry detection of yellow‐legged hornet nests, which ranged from 195 to 1331 m, with a mean value of 529 m. The traps were restricted to forest patches as this is the main hunting habitat of honey‐buzzards. ²⁸ We used a mix of water, sugar and yeast (5:2:0.06 kg, respectively) as bait. Traps were active for 7–10 days for each survey and remained inactive another 7–10 days before the next survey. In 2020, we conducted two surveys, one in the second half of July and another one in the first half of august. In 2021 and 2022, we conducted four surveys, starting the first survey in the first half of July and finishing the fourth survey in the second half of August when honey‐buzzards begin the postnuptial migration to Africa. We registered the number of hours each trap was active in the field. Traps were active for 93 572 h in total, 54 067.9 h in honey‐buzzard territories and 36 504.4 h in goshawk territories. Once collected, trapped insects were preserved in ethanol 70% until identification. To estimate the abundance of workers, we calculated the rate of capture of yellow‐legged hornet workers per hour (no. of workers/hours of activity of the trap).

2.2.4. Assessing an indicator of aerial yellow‐legged hornet nest densities in autumn

The vast majority of the full‐grown mature nests that produce reproductive individuals in autumn are aerial. ³⁴ , ⁴¹ To assess the effects on density of aerial yellow‐legged hornet nests during autumn, we used the same sampling locations of the baited traps in 2021 and 2022, and four more breeding territories discovered at the end of the season in 2022 (two honey‐buzzards, two goshawks), 21 territories in total, seven sampling locations per territory (Fig. S1). Each year in November, three proficient surveyors systematically prospected the area around each sampling location to locate the aerial nest of yellow‐legged hornet closest to the sampling location. The search area was divided in three sectors, one to be searched by each surveyor (Fig. S3). Each surveyor began the prospecting of their sector at the sampling point, moving away from it in a zig‐zag pattern to cover the whole area, using a GPS device for support. We chose November because this is when colonies are at the final stage of the biological cycle, have their maximum size and, therefore, are easier to detect. We used the distance from each sampling location to the hornet nest as a proxy of density assuming that there is a relationship between the distances measured in the field and the hornet nest densities. Thus, the higher nest densities the shorter the distances and vice versa. We decided to use this method as it was, given our logistic limitations, the only one that allowed us to obtain an indicator of density of nests from each sampling location (but not real densities), useful to test differences between honey‐buzzard and goshawk breeding territories.

2.3. Statistical analysis

We adjusted linear mixed models (LMM) over density of hornet nests attacked by honey‐buzzards per area in relation to the distance from the honey‐buzzard nest. As the density of attacked hornet nests showed an exponential distribution, we decided to perform a log–log transformation of the data. The response variable was the logarithm of the number of nests attacked per hectare, the independent variable was the logarithm of the distance from the raptor nest. We included each tagged bird, within a particular year, as a random factor.

We used linear mixed models (LMM) to test whether the honey‐buzzard travelled further from the nest to attack hornet colonies through the breeding season. The response variable was the distance from the attacked hornet nest to the honey‐buzzard nest, the independent variable was the number of days since the first chick hatched (i.e. nestling age) and the random factor was the tagged bird within a particular year. Additionally, we used linear quantile mixed models (LQMM) to test whether the maximum distances from the raptor nest to the attacked hornet nests changed along the breeding season. To do this, we performed an upper bound regression ⁴⁹ at the 0.95 quantile (τ = 0.95).

We performed LMM to study the variation in the rate of capture of yellow‐legged hornet workers along the distance from the raptor nest in honey‐buzzard territories and goshawk territories independently. The response variable was the rate of capture of yellow‐legged hornet workers per hour and we explored the effect of the distance to the raptor nest from each sampling point (0 m, 500 m and 1000 m) and the surveying period as predictors. We also compared the rate of captures at 0 m among raptors in all surveying periods, considering the species of raptor (honey‐buzzard or goshawk) as a predictor. LMM also were performed to compare the changes in the encounter distance of mature hornet nests in autumn at different distances from the raptor nest between honey‐buzzard territories and goshawk territories. To meet the assumptions of normality and homogeneity of variance of the model which allowed us to define a Gaussian distribution of error, the response variable was transformed by the logarithm (log of encounter distance of mature hornet nests) and the independent variables were the distance to the nest from each sampling location and raptor species. For all of these models, we considered the nesting territory as a random factor.

In all cases, Akaike's ⁵⁰ information criterion (AIC) was calculated for each model for the model selection process; starting from the most complex model possible to fit. A smaller AIC indicates a better‐fitting model as determined from the parsimony in the number of parameters. We used the cut‐off of AIC > 2 units to differentiate models with better explanatory power. ⁵¹ When facing similar values of AIC, we used the likelihood ratio test of nested models to test significant differences between models. In all statistical procedures we considered a level of significance of P < 0.05. All statistical analyses were performed with R software v4.2.1. ⁵² LMM models were performed with packages stats, ⁵² lme4, ⁵³ nlme ⁵⁴ and mumin. ⁵⁵ LQMM models were performed with lqmm. ⁵⁶

3. RESULTS

3.1. Yellow‐legged hornet nests attacked by honey‐buzzards

Using the pictures of the yellow‐legged hornet combs delivered to the honey‐buzzard nests by the adults sent by Spromise cameras to our computer and with the GPS data collected by the emitters before the prey deliveries (Fig. S2), we predicted the locations of 206 yellow‐legged hornet nests attacked, and we estimated the distances of the attacked hornet nests to the honey‐buzzard nests. Of the 206 yellow‐legged hornet nests attacked, we managed to locate 93 nests in the field, which allowed us to estimate the degree of destruction inflicted by honey‐buzzards. The precision of the GPS emitter (10 m) was crucial to locate the exact location of each attack in the field, but negligible to calculate the distance from honey‐buzzard nests.

Of the 93 attacked yellow‐legged hornet nests located in the field, 13 (14.0%) were aerial, located high on the branches of trees, and 80 (86.0%) were subterranean nests, founded in cavities beneath the soil or in between the gaps in the fallen leaves at the forest floor. Most of the attacked nests (89.3%) were classified as destroyed, but this percentage changed substantially with the location. For subterranean nests, this proportion reached 93.8%, whereas it represented only 61.5% of the aerial nests (Fig. 1).

Figure 1.

Bar plot showing the number of nests of Asian‐hornet attacked by honey‐buzzards and that we could locate in the field. Percentages for each category of destruction are relative to the location (subterranean or aerial). The extent of destruction inflicted on the wasp‐nest was classified into the following three categories. Almost intact: the wasp‐nest was active (workers and brood alive), two or more combs with larvae remained. Partially destroyed: the wasp‐nest was active, one comb with larvae remained. Evident signs of excavation by honey‐buzzards. Destroyed: the wasp‐nest was not active or the remaining workers staying in the burrow. No combs with brood remained.

The average most frequent distance (average mode) from raptor nests to the attacked hornet nests was 1234.7 m (Fig. S4). This supports that the scale of our experimental design (a circle of 1000‐m radius) was situated within the predator's range of action and validates the central forager behaviour of honey‐buzzards in the study area during breeding. There was a significant effect of the distance to the honey‐buzzard nest over the density of hornet nests attacked estimated in each crown segment around raptor nests (Fig. 2), showing a strong central‐forager pattern. The linear regression fitted by the LMM model (f(x) = −4.029745 − 0.000641x) predicted a significant decrease in the density of hornet nests attacked by honey‐buzzards with the distance from the honey‐buzzard nest (t = − 13.49496, df = 64, P < 0.01).

Figure 2.

Back‐transformed prediction plot of the model showing the decreasing tendency of the density of attacked Asian‐hornet nests (no. of attacks registered / ha) along the distance from the honey‐buzzard nests.

The LMM showed no significant effect of the days since hatching (nestling age) on the distance of the attacked yellow‐legged hornet nests to the honey‐buzzard nests (Fig. 3). However, the QLMM model showed a significant positive effect of the age of nestlings on the maximum foraging distance estimated using the quantile 0.95 (τ = 0.95, b ₀ = 1887.36, b ₁ = 96.77, P < 0.001) which represents the upper limit of the distribution (Fig. 3), indicating that the maximum distances of the attacked yellow‐legged hornet nests increased with the days since hatching.

Figure 3.

Linear quantile mixed effect regression estimating the change in the attack distance from honey‐buzzard nests to Asian‐hornet nests (y) as a function of the days passed since the first hatching (x). The scatterplot shows the upper bound regression (τ = 0.95) with a green line, showing a significant effect (P < 0.001) of the age of the nestlings over the maximum attack distance. Standard error was removed for clarity. Also, a linear mixed model is represented in white, with its associated standard error (grey area), showing the null effect observed over the mean attack distance.

3.2. Abundance of yellow‐legged hornet workers

The baited traps captured a total of 35 627 workers, a mean of 1298 workers in the 15 honey‐buzzard nesting territories and a mean of 2018 workers in the eight goshawk territories (Table 1).

Table 1.

Summary of captures of Asian‐hornets and hours of activity of the baited traps in nesting territories of honey‐buzzards and goshawks

Surveying period	Distance from nest (m)	Honey‐buzzard					Goshawk
		Total captured	Total trap active (h)	Mean rate	SD rate	Sample size	Total captured	Total trap active (h)	Mean rate	SD rate	Sample size
1	0	434	1361.4	0.32	0.11	9	777	1409.2	0.55	0.13	8
	500	1553	3694.8	0.42	0.16	27	1652	4138.1	0.39	0.22	24
	1000	1882	4119.2	0.48	0.25	27	1674	4086.4	0.41	0.19	24
2	0	659	2291.7	0.28	0.15	15	810	1365	0.59	0.14	8
	500	2358	6895.3	0.34	0.15	45	1634	4091.6	0.4	0.18	24
	1000	2479	6899.6	0.35	0.2	45	1393	4097.6	0.34	0.18	24
3	0	827	2658.4	0.32	0.21	15	739	1455.2	0.51	0.07	8
	500	2879	7974.9	0.37	0.2	45	1874	4362.1	0.45	0.24	24
	1000	2696	7967.1	0.35	0.18	45	1935	4364.4	0.44	0.27	24
4	0	515	1458.2	0.35	0.13	9	668	1448.7	0.46	0.07	8
	500	1539	4375.7	0.35	0.13	27	1614	4348.3	0.37	0.25	24
	1000	1661	4371.5	0.38	0.14	27	1375	4337.9	0.32	0.2	24
	Grand total/mean	19482	54067.9	0.36	0.17	336	16145	39504.4	0.44	0.18	224

We partitioned the counts by surveying periods and distances from the raptor nests.

The best‐fitting LMM model to explain the capture rate of yellow‐legged hornet workers in the baited traps in honey‐buzzard territories included a significant interacting effect between the surveying distance from the honey‐buzzard nest and the surveying period (Table S1; Fig. 4). The model predicted a mean rate of capture of 0.3041 workers h⁻¹ at 0 m from the honey‐buzzard nest. For surveying period 1, the model predicted an increase in the rate of capture over distance of 0.000116 workers × h⁻¹ × m⁻¹ (t = 3.2918; df = 289; P = 0.0011). The rate of capture decreased gradually from period 1 to 4. In period 4 the rate of capture over distance decreased −0.000017 workers × h⁻¹ × m⁻¹ (t = −0.4634; df = 289; P = 0.6434). This indicates that, in period 1, the impact of honey‐buzzards on abundance of workers was only apparent close to the raptor nest (0 m) (Fig. 4). In period 4 the effect of honey‐buzzards on hornet abundance expanded far away from the honey‐buzzard nest and was apparent at 1000 m from the raptor nest.

Figure 4.

Effect of the presence/absence of honey‐buzzards over the capture rate of workers over distance from raptor nest between sampling periods (left). Mean effect of honey‐buzzards compared to goshawks for the fourth sampling period at 0 m from the raptor nest (right). Error bars show the standard error. Defined models and selection process by Akaike information criterion is available in Table S1. Colour legend for each surveying periods is shown at the bottom.

The best‐fitting LMM model to explain the capture rate of yellow‐legged hornet workers in the baited traps in goshawk territories showed a significant interacting effect between the surveying distance from the raptor nest and the surveying period (Table S1; Fig. 4). The model predicted a mean rate of capture of 0.4965 workers h⁻¹ at 0 m from the goshawk nest. In surveying period 1, the model predicted a decrease in the rate of capture over distance of −0.000114 workers × h⁻¹ × m⁻¹ (t = −2.6152; df = 209; P = 0.0096). The decrease in the rate was more intense in surveying period 4, with a rate of capture over distance of −0.000193 workers × h⁻¹ × m⁻¹ (t = −4.4358; df = 209; P < 0.001) (Fig. 4). However, the effect among periods did not appear gradually ordered, as happened in honey‐buzzard territories. The rate of capture at 0 m was high and kept almost constant in the four surveying periods and the rate of capture at 1000 m was minimum in period 4.

The resulted LMM (rate of capture [0 m] ~ raptor species) showed a significant effect of the type of raptor over the rate of capture of yellow‐legged hornets (t = 4.357642; df = 21; P = 0.0003), predicting an increase of 0.23 workers h⁻¹ in goshawk territories compared to honey‐buzzard territories (Fig. 4).

3.3. Distances to mature yellow‐legged hornet nests in autumn

We located a total of 136 aerial nests of yellow‐legged hornet around the sampling points (at 0 m, 500 m and 1000 m from honey‐buzzard and goshawk nests) in autumn. The encounter distances ranged from 21 to 833 m in honey‐buzzard territories, and from 18 to 1013 m in goshawk territories. The distances to yellow‐legged hornet nests were similar among territories of raptor species, showing a mean value of 285.93 m in honey‐buzzard territories and of 297.21 m in goshawk territories (Table 2). The null model showed the best‐fitting LMM (Table S1) in relation to its complexity, showing no significant effect of the distances to raptor nests or the raptor species over the distances to hornet nests (Fig. S5).

Table 2.

Mean encounter distances of aerial yellow‐legged hornet nests detected in nesting territories of honey‐buzzards and goshawks

Surveying distance (m)	Honey‐buzzard			Goshawk
	Mean distance (m)	SD distance (m)	Sample size	Mean distance (m)	SD distance (m)	Sample size
0	376.86	262.25	11	141.79	66.88	10
500	279.81	278.35	33	254.33	179.13	30
1000	201.13	149.17	33	495.52	315.33	30
Grand mean	285.93	229.92	77	297.21	187.11	70

The distances also are shown partitioned by the distance from raptor nests.

4. DISCUSSION

Honey‐buzzards exhibited strong central place foraging behaviour, with intense foraging activity within the first 1234.7 m of raptor's nest, as expected for a raptor at breeding. Although previous studies had proven the consumption of yellow‐legged hornet by this raptor, ²⁵ , ²⁶ , ²⁷ , ³⁷ our study is the first to show a reduction in the population of workers in the vicinity of the honey‐buzzard nests. Despite this, we did not detect any impact on the density of full‐grown colonies of yellow‐legged hornets in autumn, an unexpected result given the impact on worker abundances. Our results suggest that despite the complexity of the ecological system controlling hornet abundances in the field, the presence of honey‐buzzards during breeding is an important driver of the abundance of yellow‐legged hornet workers which are the cause of damage to honey production and crops.

4.1. The honey‐buzzard as a central place predator

Like many other raptors during the reproductive period, ³⁷ , ⁵⁷ , ⁵⁸ honey‐buzzards were expected to behave as central place foragers, balancing the energy expenditure of hunting and transporting the prey to the nest every time. ²⁵ This defines a limiting foraging distance, beyond which hunting would become inefficient. This limiting distance marks the spatial scale of the ecosystem services provided. Raptors must carefully adapt their foraging area to the density of their preys, ensuring that the foraging distance remains energetically profitable. The ideal scenario for a foraging bird is to gather the largest amount of food during a short journey from nest, though this depends on resource availability, the energetic expenditure and nutritional quality of prey, and the individual performance. ³⁷ , ⁵⁷ , ⁵⁹ , ⁶⁰

Our results suggest that the predation pressure on yellow‐legged hornet colonies is highly dependent on distance, with honey‐buzzards locating most hornet nests within approximately the first 2500 m from the raptor nest. Beyond that point, the density of attacks decreases near to zero, as a consequence of both the reduction in foraging effort and the extensive area that would need to be covered around the nest. The optimal foraging distance for honey‐buzzards in our study area was 1234.7 m, the most frequent travelled distance from nest to prey upon yellow‐legged hornet colonies. These results support our hypothesis H.1. and validate our prediction P1.1.

The predator–prey relationship studied here is unique from an ecological and biological pest management perspective. As mentioned previously, eusocial vespids in temperate latitudes exhibit annual reproductive cycles. ⁴⁴ This means that once a colony has been preyed upon and eliminated, no other gyne is likely to establish a nest in that area until the following reproductive season. We hypothesized that this also would influence the hunting strategy of honey‐buzzards, as the predation of hornet‐nests would gradually decrease the abundance of prey in the vicinity of the raptor's nest through the reproductive season, forcing the honey‐buzzards to travel further to find more prey. In line with this hypothesis, we observed an increase of the maximum distance of attacked yellow‐legged hornet nests from the honey‐buzzard nest as the days since hatching progressed. At hatching, maximum distances were <2000 m, whereas by the time the nestlings were 50 days old, maximum distances were >7000 m, confirming our prediction P1.2. This finding indicates that within the foraging range of honey‐buzzards, the availability of hornet nests around the raptor's nest directly impacts the foraging effort required by the raptor as a response to prey depletion.

4.2. Biological control by honey‐buzzards

A crucial aspect of the honey‐buzzard's role in providing effective biological control over yellow‐legged hornet populations is its feeding strategy, as the raptor acts as a specialist predator of larvae and pupae of vespids. ²⁵ The diet of this species is unusual for a medium‐sized raptor, not only because it primarily consists of insects, but also because its prey are eusocial vespids. Other studies on birds as biological controllers of crop pests have typically focused on non‐eusocial prey, such as voles, ⁶¹ fruit‐eating birds ⁶² , ⁶³ or insects ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷ , where the elimination of an individual directly impacts overall population numbers. In the case of eusocial insect species, it is crucial that the predator impacts the colony as a whole, and this is where the honey‐buzzard demonstrates advantageous behaviour as a provider of biological control ecosystem service, with the majority of attacked hornet nests being completely destroyed.

Eusociality grants insects a more complex level of biological organization. The real reproductive unit in the population of a given species of eusocial insects is the whole colony rather than the individual. ⁶⁸ A predator consuming eusocial insect species can affect the population numbers of its prey by two processes. First, by consuming so many workers that it decreases the growth rate of the colony by affecting their ability to collect resources and feed the brood (e.g. the predation pressure of yellow‐legged hornet over bee‐hives). ⁶⁹ Second, by eliminating the colony as a unit, consuming the queen/s (eliminating the ability to produce more individuals) or the brood (in eusocial vespids, disabling the digestive function of the colony and collapsing the food‐processing chain ³¹ ). The elimination of yellow‐legged hornet colonies by honey‐buzzards falls into the second process, the consumption of the brood. As proposed by Schultner ³¹ and Bouchebti, ³² in eusocial vespid species the digestion of proteins is a specialized task performed only by larvae, which digest the food brought by their sister workers and reward them by trophallaxis with a secretion rich in amino acids, ready to be absorbed. By means of this close relationship, workers fly out to hunt avoiding the extra burden of having their stomachs full of food and only ingest easily absorbable nourishment: sugars (mostly from plant sources) and the amino acids provided by the brood. This means that the whole adult population of a vespid nest depends on the larvae. Our results suggest that most colonies attacked are highly damaged by the raptor, stripped of all offspring, which must collapse the digestive function required to keep alive the workers and the queen the time required for a next generation of larvae to be raised on time or, at least, would retard drastically the reproductive performance of the whole colony. As expected, the opposite effect was observed in goshawk territories, in which the variation of the rate of capture of workers decreased or was not affected by the distance to goshawk nests and the surveying period effect did not appear gradually ordered, indicating a noncumulative predatory effect, supporting our predictions P2.1 and P2.2. Finally, the difference in the worker abundance among raptor species at 0 m from the nest showed significant differences, which again proved our prediction P2.1. This is a remarkable result, as despite the complexity of the ecological system controlling hornet abundances in the field (e.g. production of gynes the previous season, density of embryo colonies in spring, abundance of food), the presence of honey‐buzzards during breeding seems to be an important driver of the abundance of yellow‐legged hornet workers in the ecosystem.

Against our prediction P3, the presence of honey‐buzzards did not show any effect on the density of full‐grown mature yellow‐legged hornet colonies in autumn. This result suggests that the final density of mature colonies in autumn depended on other drivers rather than the predation pressure of honey‐buzzards. It seems that, at an early stage, the ecosystem can produce an excess of embryo yellow‐legged hornet nests. These nests are mainly underground for which the predatory pressure of honey‐buzzards poses a limiting factor for their ongoing development, resulting in a quantifiable reduction of worker abundances. Subsequently, at an advanced stage, after the departure of honey‐buzzards, the success of full‐grown colonies, mainly aerial, would be conditioned by other drivers, such as intraspecific competition for resources. There are some other factors that could influence our results. (i) The difficulty of finding hornet nests only by visual means, particularly in forests dominated by nondeciduous high trees as found in our study area. We carried out the sampling in November, when the size of the full‐grown nests is large, sometimes huge, and they are visible. The canopy of Eucalyptus trees, the dominant tree in our study area, is not as dense as other nondeciduous native trees, which increases the detectability of hornet nests. However, we should consider the use of telemetry methods for future research. (ii) The search was limited to aerial nests without including the underground nests. Taking into account that the available literature supports that the vast majority of mature nests that produce reproductive individuals are aerial, ⁴¹ the impact of this second factor of bias should be limited. (iii) The indicator of nest density used in the study as proxy of nest densities was the distance from the sampling location to the nearest aerial nest. We decided to use this method as it was, given our logistic limitations, the only method that allowed us to collect an indicator of density of aerial nests in each sampling location. Despite not being a metric to directly assess yellow‐legged hornet aerial densities, we consider that it was an appropriate indicator of density to test differences between honey‐buzzard and goshawk breeding territories. New future research should address these limitations and confirm that honey‐buzzards do not affect the density of full‐grown yellow‐legged hornet nests in autumn around the raptor nests, despite the decreasing effect detected over the abundance of yellow‐legged hornet workers in summer.

4.3. Implications on the management of the invasive yellow‐legged hornet

The European Union listed yellow‐legged hornets as an invasive alien species of concern in 2014 (UE1143/2014). Our results agree with Rebollo et al. ²⁷ regarding the predation pressure of honey‐buzzards on this invasive species and demonstrate that this predation pressure influences the abundance of workers in summer. This result is of interest for the management of the exotic hornet because workers are the cause of damage to honey production and crops. Four years after colonization in 2014 in the study area, the yellow‐legged hornet became the second most important prey of honey‐buzzards. ²⁵ , ²⁷ Honey‐buzzards should be considered an ally in the fight against yellow‐legged hornet populations and should be included in the IPM programmes as a native controller of the hornet, at least on a small scale in the proximity of honey‐buzzard nests. The honey‐buzzard predatory pressure would act along with other measures such as trapping the queens in spring, manual removal of hornet nests, protection of the apiaries against yellow‐legged workers. ⁷⁰ , ⁷¹ Some studies caution about the current decrease of breeding honey‐buzzard populations in Europe. ⁷² One of the measures should be the promotion of breeding populations of the raptor in Europe by, for example, improving their forest breeding habitats. More research should be done focused on the breeding habitat preferences of this raptor as a guide for the conservation, management and, where appropriate, restoration of more suitable forest breeding habitats. The inclusion of honey‐buzzards in IPM programmes also implies that other measures to control the yellow‐legged hornet do not interfere with the ecosystem service provided by the raptor. Our main concern is the use of pesticides to eliminate yellow‐legged hornet colonies without protocols to prevent the passage of pesticides into food chains that can affect honey‐buzzards and other predators of the exotic hornet. Future long‐term research also should study the impact of honey‐buzzards on the production of reproductive individuals of yellow‐legged hornets, and the evolution of honey‐buzzards' preferences for yellow‐legged hornets and other native and exotic vespids. Additionally, the impacts of yellow‐legged hornets on the ecosystem services provided by European native vespids and on the abundance of predators specialized in the consumption of native vespids, such as honey‐buzzards, should be considered.

5. CONCLUSION

Our results provide insight into the foraging habits of the European honey‐buzzard during breeding and its impact over the yellow‐legged hornet abundances. The presence of active breeding Honey‐buzzards in the study area decreased the abundance of yellow‐legged hornet workers in the vicinity of the Honey‐buzzard nests, but its presence did not seem to affect the density of mature nests. Because yellow‐legged hornet workers are the cause of the damage to the agriculture and beekeeping sectors, honey‐buzzards during breeding are providing a quantifiable biological control ecosystem service. This study also allowed us to estimate the distance from the honey‐buzzard nest around which the honey‐buzzard provides its service. This opens an opportunity to apply this information into landscape‐management programmes which aim to increase this effect by providing sufficient breeding habitat for the honey‐buzzard. For this, future research on the honey‐buzzard's breeding habitat preferences would be necessary. These results are of interest for the management of the exotic hornet, as honey‐buzzards should be considered allies in the fight against yellow‐legged hornet populations and included in IPM programmes as a native controller of the invasive yellow‐legged hornet.

Supporting information

Data S1. Supporting Information.

DATA AVAILABILITY STATEMENT

Data are available from the Zenodo Digital Repository: https://zenodo.org/records/14020430.