Materials and Methods
The two study sites were located in the Orléans national forest (48° 00′ 28″ N, 2° 09′ 39″E) in Loiret department (45), and in the national estate of Chambord (47° 36′ 59″ N, 1° 31′ 01″ E) in the department of Loir-et-cher (41) in France. These two forests support populations of wild boar (Sus scrofa) and red deer (Cervus elaphus), as well as roe deer (Capreolus capreolus) and a small population of mouflon (Ovis orientalis) in Chambord. Installation of 2-m high fenced enclosures in these two areas has permitted to exclude (or at least to limit the number of) large ungulates but not others potentials predators such as birds or mustelids (10 cm-x-10cm mesh). In both sites, the sampling has been done by pairing fenced and unfenced plots, in forest stands with tree density and composition as similar as possible. Three types of forest stands have been chosen for this study: (i) oak regular high forests (or in conversion from coppice-with-standards to the regular high forest system) composed of pedunculate (Quercus robur) and sessile oaks (Q. petraea); (ii) pine regular high forests composed of Scots (Pinus sylvestris) or Laricio pines (P. nigra subsp. laricio); (iii) mixed oak-pine high forests. Understory was mostly constituted of bracken fern (Pteridium aquilinum), purple moor-grass (Molinia caerulea), common heather (Calluna vulgaris), associated with acidophilic species due to waterlogged winter conditions.
In 2016, four fenced/unfenced pairs had been set up (a total of 8 plots) in the national estate of Chambord only, one pair in oak high forests, and three pairs in pine high forests. In 2017, the study was extended to the Orléans forest and in each site, 9 pairs have been established (a total of 18 plots) with 3 replicates in each different forest stands presented above. For more information about each study site see Table 1. in annexes.
The two sites differed in several aspects. In the Orléans forest, we benefitted from stands whose tree composition and density are experimentally controlled and where all fenced and unfenced plots are 0.5ha in area. Enclosed plots are free of ungulate. In Chambord, enclosures are set up over much larger area (1 to 40 ha) during forest stands regeneration, in order to exclude a maximum number of large herbivores. In that case, a few ungulates may still be present in the enclosures, especially roe deer but also wild boar. We established our fenced plots in these enclosures by delimiting an area of 50 meters of radius. Finally, in order to maintain Chambord’s hunting reputation, ungulates populations are kept at a high density level in unfenced areas. In comparison, the ungulate density in the forest of Orléans is much lower to allow a balance with sylviculture activities.
Ungulates in forests are expected to impact most woodland ground nesting birds like Phylloscopus warblers and tree pipit (Anthus trivialis), and birds nesting in shrub layers like Sylvia warblers, Eurasian wren (Troglodytes troglodytes), European robin (Erithacus rubecula) and the Common nightingale (Luscinia megarhynchos) (Géroudet 1998 & 2010).These species build either opened nests with a bowl shape (Warblers), or closed nests (Eurasian wren) with a spherical shape – for the others species – (Géroudet 1998 & 2010). In order to represent these different nestling strategies (height of support and shape of nest), artificial 7-cm-diameter nests made of coco fibres and baited with two fresh Japanese quail eggs (Coturnix japonica) have been set up on each plots according to four modalities: (1) ground open-nest, (2) shrub open-nest, (3) ground close-nest, and either within the vegetation (Molinia caerulea, Rubus fruticosus and other shrubs), or in dead wood trunksand tree stumps.
Exposition to predation
Our experiment was carried out in 2016 and 2017 in Chambord forest and only in 2017 in Orléans forest. In 2016, artificial nests were exposed to predation during two successive 2-week periods in Chambord during the main bird breeding period from middle of April to middle of June (Géroudet 1998, 2010). Eggs were replaced after two weeks without changing the location of the nests, being predated or not during the first period. However, this first year of experimentation showed that nest predated during the first period tended to suffer a higher risk of predation during the second period in comparison with nests not predated during the first period, suggesting a memory effect of predators. To avoid this undesired side effect, in 2017 artificial nests were exposed to predation during a single two-week period of exposure in both forests. The average distance between nests was 14.7±3.5m in Orléans, and 16.7±4.2m in Chambord. Nests were placed so that neighbouring nests either differed in their vertical position or in their shape. (Table 1. in annexes).
In order to limit observer disturbance potentially providing sensitive cues for predators about nest location, all nests were checked only once and 14 days after set up. We noted for each nest the number of missing eggs, the state of remaining ones and the nest state. A nest has been considered as predated when it was missing, destroyed or clearly actively displaced from its original place, and/or when at least one egg was missing or damaged.
To identify nests predators and to estimate animal frequentation on each study plot during the exposition period, part of the nests were equipped with camera traps (24 Primos® Hunting, Primos Proof Cam, and model no.63056 used in Chambord in 2016 and 2017 and in Orléans in 2017 and 12 Moultrie®, GameSpy model M-80XT used in 2017 in Orléans; see Table 1 in annexes). Cameras were programmed to take sequences of three pictures separated by one second of moving animals. For each sequence, we identified the species and its abundance. All sequences within 15 min from each others were considered as a single sequence referring to the same animal (or group of animals). A calculation of the maximum of animals recorded among the sequences allowed us to estimate the abundance of each animal species during the two weeks of active camera traps. From these species abundance values per camera, we derived abundance and species richness indices at the plot scale (Table 2. in annexes). We distinguished different classes of predators: ungulates predators (wild boar and red deer), mammal predators (red fox, european badger, european wildcat, hedgehog and marten), and avian predator (great spotted woodpecker, jay, and carrion crow). Wild boar and red deer were considered separately in order to analyze their respective effect. Marten was also considered separately because it was an identified predator of our nest and so we wanted to analyze its proper effect. As Red fox, European badger, European wildcat, and Hedgehog were not clearly identified as predators of our nests, we considered them together to improve statistical power. Birds (Great spotted woodpecker, Jay, and Carrion crow) were also observed as predators of our nests, and again we combined them to reinforce statistical power.
On each plot, vegetation structure was described in three (in 2016) or four (in 2017) cardinal directions 20m away from the plot center by taking the following measures: (1) basal area of deciduous and coniferous trees using a relascope,, (2) bare ground cover including intact tree litter and bryophytes, (3) bare ground cover due to wild boar rooting, (4) cover of the herbaceous layer (all species below 50cm in height) and (5) cover of the shrub layer (vegetation between 50cm and 2m in height). All cover measurements were visually estimated (in percentage) in a radius of 3.57m (sub-plot of 40m²) (Table 2. in annexes).
Bird abundance was estimated from 10-min point counts repeated twice during the breeding season (10th May to 24th May in 2016, 26th April to 9th June in 2017), and during the four first hours after sunrise in calm weather conditions. Bird distance to plot center was measured with a laser telemeter to nearest meter and only birds within 50m from the center were kept to define bird species abundance and richness. In addition, we considered the presence with no limitation of distance of well-known egg avian predators, Eurasian jay (Garrulus glandarius), crow (Corvus corone) and great-spotted woodpeckers (Dendrocopos major) (Cramp, 1985; Baillie and al., 2002; Weidinger, 2009). We classified all recorded species according to the main nesting and foraging layers (ground/shrub versus canopy) following Baltzinger et al. (2016) and BWPi 2.0.23 (Table 3. in annexes). For analyses, we only calculated the abundance for all species without distinction, for “ground” and “shrub” nesting birds, and for “ground” and “ground/foliage” foraging birds. Species richness indices were also estimated at the plot scale (see Table 2. in annexes for details on bird variables) for all species taken together and for each foraging and nesting guild.
According to our four hypotheses, we have defined three classes of variables for our analyses: predator communities (assessed from camera traps’ surveys), bird communities (composed of variables from bird surveys), and vegetation structure (composed of variables from vegetation surveys). A total of 25 variables have been calculated and used as explained or explanatory variables depending on the ecological significance of the modelling approach (Table.2 in Annexes). To account for our sampling design characteristics, we fitted mixed models to be able to include nested “geographical” effects (nest/plot/pair/forest site). The “temporal” effects with “year” (2016 or 2017) and “period” (as nest exposures occurred at three different periods from April to June and as in 2016 the same nests were exposed to predation during two periods) have been fitted as fixed effects. According to the type of variable and its distribution, we used either linear mixed models in case of Gaussian distribution, or generalized linear mixed effect models (GLMM) for either binomial (predation rate, vegetation cover) or Poisson (abundance and richness) distributions using the glmer() function of the R lme4 package (R Development Core Team- 3.4.0). In order to correct the abundance of predators at plot scale by the number of camera set up on each plot, the glmer() function’s argument offset() with the log(camera number). For each variable that had to be explained, most influential explanatory variables were identified using a forward variable-selection approach based on Akaike information criterion corrected for small sample size (AICc). Closely related variables (eg abundance/richness of predators) were never put simultaneously in the same model. Error risk α is always determined at 0.05%. Throughout the rest of this report, results are always presented as mean ± standard error.
Data analysis primarily aimed at identifying effect of ungulate density on nest predation and bird communities. From this standpoint, we investigated (1) how predation was influenced by fencing and ungulate abundance, (2) how predator community (mammal and avian) was related to fencing or ungulate abundance, (3) how the bird communities were related to predation risk, predator abundances, vegetation features, fencing and study site (4) which variables explained spatio-temporal variation in nest predation rate among those related to plot type (fenced/unfenced), ungulates and other egg predators, bird communities and plot vegetal features.
Throughout the rest of the document, final models in terms of selected variables and their results are detailed in annexes in R language.
Table of contents :
2. Materials and Methods
2.1. Study area
2.2. Experimental design
2.2.1. Artificial nests
2.2.2. Exposition to predation
2.2.3. Camera traps
2.2.3. Vegetation survey
2.2.4. Birds surveys
2.3. Data analysis
3.1. Effect of fencing on predation
3.2. Effect of fencing on nest predator abundances
3.3. Effect of fencing on bird communities
3.4. Nest predation risk
4.1. Nest predation risk related to fences
4.2. Impacts of ungulates on nest predation
4.2.1. Direct impact: nest predation
4.2.2. Indirect impact: attraction of other predators
4.2.3. Indirect impact: modification of environment
4.3. Impacts of ungulates on bird communities
4.4. Influence of nest height in predation rate