The potential role of scavengers in spreading African swine fever among wild boar.

Understanding the transmission patterns of African swine fever (ASF) among wild boar (Sus scrofa) is an issue of major interest, especially in the wake of the current ASF epidemic. Given the high stability of ASF-virus, there is concern about scavengers spreading infectious carcass material in the environment. Here, we describe scavenging activities on 32 wild boar carcasses in their natural habitat in Germany. Using digital cameras, we detected 22 vertebrates at the study sites, thereof two mammal and three bird species scavenging. The most frequently detected species was the raccoon dog Nyctereutes procyonoides (44% of all visits). Raccoon dogs, red foxes (Vulpes vulpes), and buzzards (Buteo buteo) scavenged in the warm and the cold season, while ravens (Corvus corax) and white-tailed eagles (Haliaeetus albicilla) scavenged only in the cold season. In summer, however, insects removed most of the carcass biomass. Although most of the material was consumed on the spot, foxes, raccoon dogs and ravens left the study sites in rare cases with a small piece of meat in their mouths or beaks. We conclude that scavengers represent a minor risk factor for spreading ASF, but may contribute to reducing local virus persistence by metabolizing infected carcasses.


Results
Animal species. All carcasses were subject to depredation by scavengers or decomposition by insects. Both mammals and birds were present at all nine sites (Fig. 1). Piglets 6 /7 and 8 were not tied securely enough, so they were dragged away by raccoon dogs. Carcass 1 sank unintentionally under water for three months.
The 13 cameras yielded a total of 122,160 images. The nine selected cameras yielded a total of 67,967 images that could be evaluated (i.e. at least one identifiable animal was displayed), thereof 76% (n = 51,718) in summer/ autumn and 24% (n = 16,249) in winter/spring. According to the previously defined time interval of eight minutes between two visits, the images that displayed at least one identifiable animal (n = 67,967) were assigned to a total of 6,153 visits.
During day and night, the study sites were meeting points of interspecies encounters. A total of 243 images displayed two species simultaneously, thereof 165 images of different birds (139 buzzard and raven; 23 raven and eagle; 3 buzzard and hawk); 67 images of different mammals (46 red fox and raccoon dog; 9 red fox and marten; 4 raccoon dog and polecat; 3 raccoon dog and marten; 3 raccoon dog and wild boar; 2 red fox and raccoon) and 11 images of a mammal and a bird (6 wild boar and a buzzard; 5 red fox and raven) ( Supplementary Fig. S1).
Two mammals (foxes and raccoon dogs) and three birds (ravens, buzzards and eagles), were clearly observed scavenging on carcasses of different decomposition stages. Foxes and raccoon dogs were observed scavenging at all sites in both, the warm and the cold season. Buzzards and ravens were observed scavenging at all sites, but ravens only in the cold season. White-tailed eagles were observed only in the cold season on sites 1-3. Raccoons, different species of Mustelidae, water voles and several species of passerine birds were also observed for short intervals directly on the carcasses. It was not clear, however, if they had eaten carcass material.
Raccoon dogs and foxes showed interest in the carcasses from day 1 after exposure until the stage when they had completely decomposed ( Supplementary Fig. S2). Both species were observed vigorously tearing at carcasses and pulling intestines out of them ( Supplementary Fig. S3). In advanced stages of decomposition, they were also observed ripping soft parts and skins into pieces and carrying them away (Fig. 2). Supplementary Video S6 captures foxes scavenging and pulling on carcass 12 from day 25 after exposure. Two times, ravens left the visual field of the camera with a small piece of meat in their beaks (Fig. 2). Especially during summer, raccoon dogs were observed scavenging together with their offspring in groups of up to eight individuals ( Supplementary Fig. S2). Foxes scavenged individually or in pairs ( Supplementary Fig. S2), raccoons were observed solitary or in groups of up to three animals ( Supplementary Fig. S4). Buzzards fed and fought in pairs and ravens assembled in feeding groups of up to 14 animals ( Supplementary Fig. S2).
Hypothesis 1: The first animals that detect wild boar carcasses are significantly more frequently birds (>50%) than mammals. In  animals that first detected the carcass were birds, namely buzzards (14), ravens (2), nuthatch (1) and starling (1). In 11 cases (38%), mammals detected the carcass, namely raccoon dogs (6), wild boar (3) and foxes (2). The difference between birds and mammals was not statistically significant (95% CI: 0.45-1.00; p = 0.133, one-tailed test). However, when the data were separately stratified by season and site visibility, in winter/spring or in forest clearings, it was mostly birds that first detected the carcasses (birds 87% vs. mammals 13%), while in summer/ autumn and in closed forests, it was mostly mammals (mammals 64% vs. birds 36%; p = 0.008 for both season and site visibility). Multivariable analysis also revealed that in forest clearings and winter/spring, the chance that the carcass-detecting animal was a bird was higher ( Table 2). Hypothesis 2: Carcass detection time is shorter for birds than for mammals. The detection time ranged between 0 and 22 days in summer/autumn (median 1.5) and 0 and 16 days in winter/spring (median 6; p = 0.844). No statistically significant difference was found between mammals and birds. Mammals (n = 11) detected the carcasses between day 0 and 22 (median 4) and birds (n = 18) between day 0 and 17 (median 3; p = 0.58). Multivariable analysis with adjustment of all covariates revealed that neither body weight nor carcass type, season of exposure or site visibility had a significant influence on the carcass detection time. In closed forests, the time ranged between 0 and 22 days (median 3) and in forest clearings between 0 and 17 days (median 4; p = 0.93).
Hypothesis 3: Persistence time of the carcass depends on the body weight, carcass type, site visibility, and exposure season. Depending on the body weight and exposure season, it took between 4 days (young female wild boar exposed in summer) and 12 weeks (adult male boar exposed in autumn) until skeletonization was complete, i.e. only bones and desiccated skin were left (Table 3). For statistical analysis, carcass 1 was excluded as it had drowned in water and its decay was not comparable to that of the other carcasses. Median persistence time varied between 8 days in summer/autumn and 37 days in winter/spring (p < 0.001; Fig. 3). Univariable analyses showed no significant association with respect to body weight, carcass type and site visibility. Multivariable analyses with inclusion of exposure season, body weight, carcass type and site visibility revealed that the carcass persistence time was significantly longer in winter/spring than in summer/ autumn (adjusted Hazard Ratio HR = 0.04, 95% CI: 0.007-0.218, p < 0.001). It became also evident that heavier carcasses persisted longer than lighter ones. Cox regression with stepwise backward variable selection showed that season and body weight were the only remaining factors that explained the differences in carcass persistence time ( Table 2).   (Table 1). Most visits (n = 2,914) took place in the first two weeks (Fig. 4a). Half of the mammal visits took place in the first 14 days, half of the bird visits in the first 9 days (Fig. 4b).
Both site visibility and carcass type had also a significant influence on the number of visits: Carcasses exposed in forest clearings were visited more often than carcasses in closed forests (4,044 visits; 69% vs. 1,784; 31%; p < 0.001). Both mammals and birds were present more often in forest clearings (2,735; 64% of all mammal visits and 1,309; 84% of all bird visits) than in closed forests (1,532; 36% and 252; 16%).
After adjusting for the effects of all other covariates, significant effects on the number of visits were found for (i) the taxonomic group (the chance to be visited by a mammal was about 1.7 times higher than to be visited by a bird), (ii) carcass type (the chance for a whole adult to be visited was 1.8 times higher than for an eviscerated adult), (iii) time since exposure (the chance for a visit was highest in the first two weeks), and (iv) site visibility (the chance for a visit in a forest clearing was 1.6 times higher than in a closed forest) ( Table 4). Random effects (individual carcass number and site number) had no significant influence on the number of visits (Table 4).

Discussion
In our study, all wild carnivores known to be endemically present were observed at least once, including the raccoon dog and the red fox, which is the second largest carnivore in Germany after the wolf, Canis lupus, which is not a common resident in the study area. Raccoon dogs and red foxes are flexible omnivores and opportunistic scavengers with a wide-ranging diet 46,47 . We observed them scavenging in both the warm and cold season, when

Carcass detection by a bird
Carcass persistence time  Table 3. Wild boar carcasses exposed to scavengers. www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ insects and microbes are less active, which is in line with previous studies 48,49 . Mammals mostly scavenged at night and birds during the day. This was expected, since the observed mammals are mainly nocturnal and the birds diurnal.
We found that birds were slightly faster in detecting carcasses than mammals in forest clearings and in the cold season. Birds have a number of adaptations that give them an advantage over mammals when it comes to carcass discovery, including a panoramic view, a keen sense of vision and social information transfer 19 . Ravens are usually the first species to arrive at large carcasses 50,51 and share knowledge of feeding opportunities 52 . However, when exposing the carcasses in the field, we could not avoid leaving acoustic, olfactory and visual cues that could have made carcass detection more efficient to particular species. In addition, we might have unintentionally attracted specific scavenger species to the study sites by our peculiar behaviour. Since buzzards and ravens are communal feeders, their presence may indicate the presence of a carcass to mammal scavengers in the vicinity by their flying activities and vocalizations 21 . In our study, the largest groups that assembled at the carcasses were formed by ravens, with up to 14 individuals observed simultaneously. Previous studies found more than 20 ravens near a carcass at the same time 53 . Other studies also indicate that crows are responsible for a considerable proportion of carrion removal 26 .
In our study, the carcass detection time ranged between 0 and 22 days. We found no significant relationship between the detection time and season or the location of the carcass, namely site visibility. This is not in line with previous studies, where carcasses placed in open areas were detected and consumed at faster rates than carcasses under closed forest canopies 33 . This discrepancy might be due to the fact that on two closed forest sites, we placed several carcasses consecutively. Scavengers may thus have become habituated to this regular feeding source equally well in forest clearings and under closed forest canopies.
Although the winter was relatively mild, in the cold season the carcasses persisted several weeks or even months longer than in the warm season. Previous studies have already shown that the persistence time of a carcass is influenced by a complex interplay between weather 38,54 , type of habitat (exposure to sun or moisture, soil type etc.) 40 , presence of scavengers 55 , body size 40,41 and integrity of the body surface 42 . Regarding the latter, we exposed 12 whole carcasses of adult wild boar. Before scavengers gained access to the inner organs and could remove larger amounts of tissues, they had to perforate the skin. The skin of wild boar, especially of adults, is much thicker than the skin of domestic pigs and seems to be particularly difficult to open. In winter, buzzards and ravens were observed picking on the torso for hours before they were able to open the first hole in the skin. Overall, in the cold season, this process was slow and tissues were preserved for several weeks. It is therefore conceivable that the carcass of wild boar that succumbed to ASF in late autumn might persist in the environment until next spring, serving then as a new source of infection. This risk was lower if the animal had died in the warm season, as the metabolization of its biomass would be significantly accelerated by invertebrates. The fact that warmer temperatures result in higher insect activity and thus in faster carcass decomposition was expected and has previously been described 56,57 .
It became obvious that insects versus vertebrates on the one hand and birds versus mammals on the other hand exhibited seasonal variability in their presence. Mammals scavenged in both, the warm and the cold season, while ravens and eagles scavenged only in the cold season. Although buzzards, ravens and eagles are resident year-round in the study area, they were mainly (eagles exclusively) observed in winter. This finding is in line with previous studies on the foraging strategy of white-tailed eagles, which documented that foraging flight duration was lowest in spring and increased over summer and autumn towards winter 16,18 . Possibly, their preferred prey (birds, fish, and small mammals) 58 was not available in winter, so that they fed on carrion instead.
The most frequent visitors, however, were mammal scavengers, namely raccoon dogs and foxes. Not surprisingly, the season influenced their number of visits, but in the opposite way, we had expected. We had anticipated that in the cold season, when food resources are scarce, scavengers would supplement their diet more often with carrion than in the warm season. This was not the case. One reason might be that spring and summer is breeding season, during which mammalian females need an increased energy intake to cope with lactation. Especially raccoon dogs have been observed in family groups at the study sites for longer periods, probably due to the nutritional needs of the offspring. Another reason might be that we conducted the study in a region with a high population density of ungulates and a wide variety of anthropogenic food sources. Coupled with relatively mild temperatures, this may have been responsible for the low number of visits during the cold season. However, there might also be a caveat associated with this finding, since the number of carcasses exposed in summer was relatively high (n = 10).
We could not recognize a regular pattern in the sequence of scavengers or decomposition stages. The main reason was the unpredictable presence and feeding activity of scavengers. The strong influence of scavenging on the decay rate of carcasses has been noted previously 59,60 . The effects of scavenger activity were so manifold, that our idea of determining a "standard scheme of decay" for estimating the time of death of a wild boar found dead was not practicable. Also, many different individual characteristics (gender, age; gutted, wounded or with intact body surface etc.) and environmental conditions (weather, soil composition etc.) made it impossible to draw a scheme of decay or derive general rules on the "standard" scavenging behaviour towards a wild boar carcass.
Raccoons, badgers, water voles and several insectivorous birds were also observed at the study sites ( Supplementary Fig. S4). We could not distinguish if they were attracted by the carcasses or by prey insects that assembled to consume organic material at the decomposition islands. Great tits, for example, have a welldeveloped sense of smell 61 and might profit from carrion opportunistically through predation on necrophagous insects. Visiting, but non-scavenging species (deer, squirrels, etc.) seemed to have had no influence on the fate of the carcasses. We identified 22 different species on the images. It was not possible, however, to identify individual animals. It is therefore possible that individual animals visited carcasses repeatedly. However, except for sites one and two, the distance between the study sites was several kilometers. It is therefore unlikely that the same animals visited different sites. www.nature.com/scientificreports www.nature.com/scientificreports/ In this study, carcasses were mainly consumed at the site of exposure, except for three piglets that were dragged away. Since we tried to prevent dispersal by tying the carcasses with a rope to a pole, we cannot rule out that dragging away small carcasses is a common scavenging practice. The study design probably affected not only the number of dispersal events, but also the detectability of carcasses. For instance, under natural conditions a carcass (at least the small ones) may have been pulled into a bush, making it possibly more difficult for other animals to find or get access to it. Foxes and raccoon dogs were observed vigorously tearing the intestines out of the abdomens of carcasses of adult animals and ripping soft parts into pieces and carrying them away. In addition, ravens left the visual field of the cameras with small pieces of meat in their beaks. Red foxes and other scavengers are known for hoarding surplus food for future consumption 62,63 . We cannot quantify the amount of carcass tissue or the number of pieces that these individuals scattered around to consume them on a safe spot or to feed their young. However, from what we could see on the images, the amount of dispersed material is probably negligible. Furthermore, we assume that bones or other carcass material might be scattered over a radius representing the home range of the animal at maximum. In Mecklenburg-Western Pomerania, the study region, the average annual home range sizes of male and female raccoon dogs and red foxes are estimated between 161 and 177 ha respectively 64 . This area is smaller than the home range of wild boar family groups (the relevant species for ASF), which has been reported to vary from 29 to 685 ha and up to 3,480 ha (after hunting) 65 . Therefore, wild boar in the incubation period would anyway be more effective in spreading the disease by daily movements than scavengers by dispersing contaminated material.
During the warm season, the risk that scavengers spread ASFV might be even lower, since most carcass material is not consumed by mammals or birds, but rapidly metabolized by microbes and insects. In summer, whole carcasses were almost completely decomposed within days by carrion flies, maggots and beetles. This allowed significant amounts of nutrients to stay at the site and enter the soil, thereby forming so-called carcass decomposition islands. The relationships between these islands, decomposition, as well as insect and scavenger activity have been well described 55 .
This study has been designed as an observational study. As all experimental studies in natural settings, the results should be interpreted with caution. The sample size was limited to 32 carcasses and they had many different characteristics. Hence, the statistical procedures for analyzing the data are intended to be exploratory.
Our results show that especially in winter, raccoon dogs, foxes, buzzards and ravens play an important role in the decomposition process of wild boar carcasses, while in summer, most of the carcass biomass is rapidly decomposed by invertebrates. Although we cannot rule out that in rare cases, scavengers might disperse small pieces of infectious material in the near surroundings of a carcass in an ASF-affected region, it seems unlikely that this could have a major impact on the spread of ASF in an affected region. Previous studies suggest that scavengers may even contribute to reducing the transmission potential by removing infected material from the environment as long as they are no competent hosts for the pathogens exposed to by scavenging 66,67 . Upon ingestion, ASFV is extremely unlikely to remain infectious after passaging through the intestinal tract of a vertebrate (Sandra Blome, personal communication). Moreover, ASFV replicates in cells of the mononuclear phagocyte system of suids 68 and certainly not on the body surface of a scavenger. In addition, mammals and birds usually groom themselves, when their fur or feathers get dirty with blood or other body fluids. Regarding the larvae of necrophagous insects, they do not play a relevant role as mechanical vectors for ASFV, but even seem to have an inactivating effect 69 .
In conclusion, we do not think that scavengers are epidemiologically relevant risk factors or that reducing the population of scavengers in ASF-affected regions is likely to have an effect on disease control. On the contrary, scavengers are efficient in removing wild boar carcasses and may thereby contribute to reducing the risk of virus persistence in the environment.

Methods
Study design. Thirty-two wild boar carcasses were exposed on nine sites around the town of Greifswald, northeast Germany (54°6 N, 13°23 E). The landscape is dominated by agriculture, forestry and a low number of human settlements. Ungulate density including wild boar, red deer, fallow deer (Dama dama) and roe deer is high. The sites were located away from paths in five different mixed coniferous/ deciduous forests dominated by oak (Quercus robur), alder (Alnus glutinosa) and beech (Fagus sylvatica). The forests are surrounded by crop production fields, where mainly wheat, maize, and oil-seed rape are grown. The climate is influenced by the Baltic Sea. The cold season (December to March) usually covers periods with snow and temperatures below zero. During the study period, monthly mean temperatures varied between 7.3 °C, 6.7 °C, 0.8 °C, 3.4 °C, 4.3 °C, 7.8 °C and 13.9 °C between November 2015 and May 2016, respectively. The chosen locations were particularly inaccessible away from paths. Seven sites were located in forest clearings, and two sites underneath closed canopies, i.e. shady and windless. The understorey was sparse and relatively similar across all the study sites. Only on sites 2 and 3, the herbaceous layer was dense. These sites were mown with a scythe when necessary for a free view on the carcass.
The study was conducted from 27 October 2015 until 27 October 2016, i.e. in a total of 367 days and nights. During the whole study period, no hunting took place in the surroundings of approx. 1 km of each study site. Seven carcasses (12-18) were exposed in spring, ten (19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29) in summer, four in autumn (1 and 30-32) and eight (2)(3)(4)(5)(6)(7)(8)(9)(10)(11) in winter (Table 3). Fourteen carcasses were exposed under closed canopies (sites 2 and 9) and 15 in forest clearings (sites 1 and 3-8). The carcasses of piglets 4 and 5, 6 and 7 as well as 27 and 28, were exposed on the same day and close together. They were so small that each pair was counted only as a single carcass, resulting in 29 carcasses included in the statistical analysis, thereof 11 adult males, 10 adult females and 8 piglets (≤20 kg). On six sites (1 and 4-8), we placed one carcass, and on sites 2, 3 and 9, we consecutively placed eight, nine and six carcasses, respectively. All carcasses were monitored until completely consumed by scavengers or decomposed by invertebrates and only bones and skins were left over. The carcasses were either placed on the study site immediately after they had been obtained or were cooled until they could be brought to the site of exposure. They were www.nature.com/scientificreports www.nature.com/scientificreports/ placed directly on the forest ground so that terrestrial animals, birds and invertebrates had unrestricted access. To avoid that the carcasses were pulled outside the visual field of the cameras, they were tied up to a pole with a rope.
Each carcass was monitored either by an infrared heat-or a motion-sensitive digital camera. We used five different game camera models, namely Seissiger Special-Cam 3 Classic (Anton Seissiger GmbH, Würzburg, Germany), Maginon WK 3 HD (Supra, Kaiserslautern, Germany), Dörr Snapshot UV555 (Dörr GmbH, Neu-Ulm, Germany), Moultry A5 and Moultry I40 (Moultrie Alabaster, USA). The cameras were on standby mode during 24 hours a day and set to take one photo every 1 to 10 s if activated. They were installed on trees in a distance of 4 to 8 meters to the carcass and a height of 1 to 2.2 m above the ground. All cameras were directed and re-arranged as necessary to keep the carcasses in view and to capture all movements and animal activities in a radius of approx. 4 to 10 m around the carcasses. Date, time, and temperature were recorded automatically on each picture. In winter 2015/2016, we installed one camera at each site. To observe not only the immediate proximity, but also the surroundings of the carcasses and to capture approaching animals from all angles, we installed two additional cameras at sites 2 and 3 in summer 2016. However, as the focus of the present study was to quantify scavenger activity on the carcass itself, we used only data from one camera at each site, so that double counts of animal visits were excluded. Every two to fourteen days, the carcasses and the cameras were inspected. During each visit, pictures were taken with a handheld digital camera (Canon Power Shot A 3400) to document the status of carcass decomposition.
The Veterinary Authority of the district of Vorpommern-Greifswald was duely informed about the study and the Department of Forestry of Greifswald as the competent authority participated in it. Since no live animals were used and the wild boar carcasses had been purchased from local hunters, no ethics approval was necessary.
Data analysis. Depending on the characteristics and spatiotemporal conditions, we divided the study sites, carcasses and visiting animals into categories. Regarding visibility from above, study sites were divided into two categories, depending on the density of the overhead canopy of branches and foliage (closed forest; forest clearing). With respect to body weight, carcasses were assigned to three categories, i.e. piglets (≤20 kg), young (21-60 kg) and adult pigs (>60 kg). With regard to the age of the animal and its status after death, the carcasses were assigned to three 'carcass types' (adult whole, adult eviscerated or piglet). Stratified by taxonomic group (mammal, bird), usual food sources and whether the species was observed scavenging or not, visiting animals were assigned to five categories (scavenging bird, potential scavenging bird, mammal scavenger, potential mammal scavenger and other mammal). Dates of exposure were assigned to the respective quarter of the year, namely summer (June-August), autumn (September-November), winter (December-February) and spring (March-May). For statistical analysis, only two seasons were distinguished, namely the warm (summer/autumn) and the cold (winter/spring) season.
We counted any presence of an animal on the visual field of the cameras as a "visit". All images made of the same species in a temporal context were assigned to a single visit: If the time interval between two images was longer than eight minutes or a different species was recorded, the image was assigned to a new visit, if not, it was assigned to the same visit. If two different species were recorded at the same time, the visit was assigned to both species. Images were classified into primary visits (first visit of a species to the carcass), first direct contacts (first physical contact of that species with the carcass) and secondary visits (all following visits, regardless of involving contacts or not). Since animals might visit the study sites just by chance, i.e. without being aware of the carcass, we regarded as "carcass detection" the first visit with physical contact with the carcass (sitting on the carcass or touching it with the snout or beak). The time (days) elapsed between carcass placement and the first physical contact with the carcass was interpreted as the "carcass detection time". "Scavenging" events were derived from the animal species involved, the observed behaviour (physical contact with the snout or beak and a clear attitude of consumption), the duration of contact and a change in the position of the carcass. As "carcass persistence time" we defined the time that elapsed from exposure to complete skeletonization (only fur and bare bones left).
To test our hypotheses, we chose the following variables: First detection of a carcass (Hypothesis 1): taxonomic group of detecting animal (bird, mammal). Carcass detection time (Hypothesis 2): taxonomic group of detecting animal, exposure season and site visibility. Persistence time of a carcass (Hypothesis 3): carcass characteristics (body weight, carcass type), site visibility and exposure season. Number of visits (Hypothesis 4): species of visiting animal, time since exposure, exposure season (as an indicator for food availability), body weight, carcass type, and site visibility. Since several carcasses were exposed at the same study sites, individual carcasses clustered on the same sites were included as random variables.
To test Hypothesis 1 (the first animals that detect wild boar carcasses are significantly more frequently birds than mammals), we used the exact binomial test for univariable analysis. To prove the influence of covariates (exposure season, site visibility, carcass type, and body weight), we tested them individually against the frequencies of birds and mammals (univariable; Fisher´s exact test). Finally, we used a logistic regression model with stepwise backward variable selection using the Akaike Information Criterion (AIC) approach to select the variables that best describe the data.
To test if the carcass detection time was shorter for birds than for mammals (Hypothesis 2), we used the log-rank test and generated Kaplan-Meier plots for univariable testing. To assess the influence of covariates (exposure season, site visibility, carcass type, and body weight), we used a Cox regression model with stepwise backward variable selection for multivariable testing.
To test Hypothesis 3 (factors that shape the persistence time of a carcass), we used the log-rank test for univariable testing. To assess the influence of covariates, we used a Cox regression model with stepwise backward variable selection for multivariable testing.
To test Hypothesis 4 (influence of the type of visiting species, time since exposure, season, body weight, carcass type, and site visibility on the number of visits), we first analyzed the distribution of the data. Since they are similar to a Poisson distribution, we first used a Poisson model with and without random effects. This model was