The older the richer : significant increase in breeding bird diversity along an age gradient of different coppiced woods

Forest structural complexity could be a good predictor of overall species diversity. Since tree harvesting has a negative effect on forest structure, it is important to analyse the effects of this disturbance on sensitive groups, as forest birds. In this study, we aimed to shed light on this aspect by analysing a set of univariate metrics in bird communities breeding in three coppiced forest habitats (coppiced of chestnut, coppiced of Turkey oak and high forest of beech) along a gradient in age classes. We hypothesised that, with increasing forest age, (i) breeding bird communities will progressively increase in diversity and, (ii) due to higher habitat heterogeneity due to coppicing, a higher species turnover in the first age classes could appear. In each forest habitat, all the metrics significantly increased, from recently coppiced to more mature forests, due to progressively higher availability of resources and niches along the gradient. When comparing paired forest habitats, abundance and richness were significantly different only in the two oldest age classes, highlighting that responses to different tree composition were more marked in the mature phase. In all forest habitats, species turnover (βw diversity) decreased progressively along the age gradient and was highest in the youngest age classes where many vegetation layers were present. Due to different coppice management practices, growth regime and consequent habitat heterogeneity, chestnuts showed a different pattern when compared to other forest habitats, with an increase in species turnover (βw diversity) at intermediate level. With increasing age of the forests, all the diversity metrics increased and species turnover decreased, highlighting the role of older forests as strategic habitats for highly structured bird communities.


Introduction
Forests are universally recognised as a mosaic of complex and multifunctional ecosystems with intrinsic value.Therefore, forest management cannot be based only on the principles of the market economy but also on those of biodiversity conservation and ecosystem services (Lindenmayer et al., 2000;Siitonen, 2001;Pardini et al., 2005Pardini et al., , 2010)).
The structural complexity of a forest seems to be a good predictor of overall species diversity (Annand and Thompson, 1997;Dìaz et al., 2005).Accordingly, analysis of forest structures is known as a reliable criterion with which to assess the conservation value of forest stages (Doyon et al., 2005).Moreover, the structural changes of the forests have implications on their ecological functions, by alteration of the microclimate, the production of food resources and the capacity to provide shelter or nesting sites to animals (Wiens, 1989;Chapin III et al., 2000).
Forest birds, for their easy detectability and high sensitivity to human-induced environmental stress, are an important group for analysing the effects of forest ecosystem transformations (Blondel, 1975(Blondel, , 1981a;;Wiens, 1989;Noss, 1990;Villard, 1998).Moreover, these animals depend very heavily on vegetation structure and their populations can be deeply affected by forest cutting (Ferry and Frochot, 1974;Canterbury et al., 2000;Jobes et al., 2004), one of the major an-L.Mentil et al.: Bird diversity in different aged coppiced woods thropogenic threats recognised by the IUCN (International Union for Conservation of Nature, code 5.3 "Logging and wood harvesting"; review in Brawn et al., 2001;Battisti et al., 2016).
Tree harvesting has a negative effect on forest bird populations at different levels.Its intensity depends on spatial and soil features, as well as on the forest type (Donovan et al., 1997;Villard et al., 1999;Austen et al., 2001).Therefore, it is important to analyse the ecological effects of forest cuts, especially when these occur with high frequency and intensity as in the case of coppicing.Nevertheless, in southern Europe studies in this direction are few and rarely analyse the structural evolution of the differently characterised coppiced woods along age classes (Ciancio et al., 2006;Gil-Tena et al., 2007;Nascimbene et al., 2007;Torras and Saura, 2008;Spinelli et al., 2010).
The aim of this paper is to compare the structure of the bird communities in three different types of managed forests (Turkey oak coppicing, chestnut coppicing and beech high forest) belonging to different age classes in a regional nature reserve.Since these habitat types increase their structural complexity, moving from recently coppiced to high forest (and since structural complexity is a strong driving forces of resource and niche availability; Wiens, 1989), we hypothesised that breeding bird communities will progressively increase their univariate metrics of diversity, such as species abundance, richness and diversity.Moreover, since immediately after cutting, forest habitats become more heterogeneous, with the presence of open habitats and shrub vegetation, we hypothesised a higher species turnover (calculated with a β diversity metric) in the first age classes when compared to the more mature ones.

Study area
The study was carried out inside the Bracciano-Martignano Regional Park (provinces of Rome and Viterbo, central Italy, Lazio; Fig. 1), in the northern sector of the Sabatini Mountains (size: 3000 ha, geographic coordinates: 42 • 08 53 N, 12 • 09 55 E).This area include volcanic hills, with many pyroclastic cinder cones (e.g.M. Calvi, M. Termine, M. Rocca Romana) overlooking the caldera depressions, the largest of which harbours the lakes of Bracciano and Martignano.The altitude ranges between 164 m (Lake Bracciano) and 612 m a.s.l.(top of Mount Rocca Romana).The vegetation is dominated by broadleaf deciduous woodland, from oak to beech stands, with anthropogenic chestnut orchards (Sollevanti, 1983).
Forest habitats differ in structural traits due to different practices of coppice management: in particular, in chestnuts and Turkey oak woods traditional practices of coppicing were carried out with a specific time frequency for cutting (about 15-18 years).In contrast, beech forest were managed as high forest and practices are completely different (Ciancio et al., 2006).In southern Europe, high forest is a form of government characterised by a turn of about 90 years, followed by the birth of seedlings (details in Table 1).
For each forest habitat, we selected four age classes following the dynamic steps of their structure and local history of coppicing (Pregitzer and Euskirchen, 2004;Scarfò, 2012): the first three age classes belong to different succeeding steps of cutting, while the fourth step (an unmanaged wood) has been considered a control (i.e.woods with the oldest age when compared to coppicing turn; Scarfò, 2012; Table 1).

Protocol
Breeding bird communities was studied in the 2014 spring season (from March to May) with the point count method (Bibby et al., 2000).This method is particularly suitable for studies in patchy landscapes and in the spring period when species show territorial behaviour, so they can be easily de-  tected (Blondel, 1981b;Sutherland, 2006).We used a 4 (age classes) × 3 (forest habitats) factorial design.
In each age class of each forest habitats, we located six point counts.Sampling points were randomly selected using a random number generator: http://stattrek.com/statistics/random-number-generator.aspx, last access: April 2014) and subsequently identifying the geographical coordinates using the free software Google Earth and a GPS Garmin E-trex on the field.We randomly located point counts (n = 72).To evaluate the efficiency of our sampling, we built 12 data matrices, one for each site, corresponding to each species and treatment and then conducted a rarefaction analysis by building rarefaction curves with 95 % confidence intervals calculated after 9999 bootstraps (PAST 3.0 statistical software).Curves for age classes were grouped for each forest habitat.We observed rapid cumulating patterns in all treatments (outputs in Supplement Sect.S1): therefore we assumed that our sampling effort was representative of breeding bird assemblages, at least for common species.
Within each point count we recorded any individual bird seen and/or heard (song, call, drumming, alarm display), during a standardised fixed time (5 min), within a radius of 50 m.The use of this sampling distance was considered reliable for reducing the bias of detectability between mature and recently cut woodlands, where the probability of contacting birds is much broader and less sound absorbing because of dense vegetation (Sutherland, 2006).
In each point count we carried out two sampling sessions (session I: March, from 7 to 25; II: from 1 to 11 May 2014; total time effort: 720 min).We recorded birds in the morning from dawn to approximately 10:00.The distance among point counts remained at least 250 m to avoid counting the same individuals several times (pseudoreplication bias; see Rossi de Gasperis et al., 2016).For each species, in each point count, the highest value recorded in the two sampling sessions was considered valid (Blondel, 1975;Malavasi et al., 2009).
We used Nikon Monarch binoculars 12 × 42 DCF (roof prism with centre focus) and recorded calls using an iPhone 4s with a mini-microphone and the free software iTalk.

Data analyses
For each age class of each forest habitat, we obtained the species-specific abundance (number of individuals, n) and their relative frequency (Fr = n/N, where N is the total number of individuals sampled).At the community level, we calculated the following univariate metrics of diversity: mean total abundance (N mean ), species richness (S), mean species richness (S mean ) and Shannon-Wiener diversity index (as H = − Fr ln Fr; Shannon and Weaver, 1963;review in Magurran, 2004;Magurran and McGill, 2011).
To assess the level of intra-turnover of species inside the bird communities, we calculated the β w diversity index (Whittaker, 1972) as follows: where γ is the total number of species recorded in the community (i.e. the γ diversity, here corresponding to the S value) and α is the averaged species richness (i.e. the α diversity, here corresponding to the S mean value).
To test if there are significant differences among mean values among multiple paired samples (i.e.samples with the same number of replicates, n = 6), we carried out the non-parametric Friedman test.To test the differences between paired mean values, we carried out the non-parametric Wilcoxon test (Dytham, 2011).Alfa was set at 0.05.We used the free software PAST (Hammer et al., 2001).

Results
In total, we recorded 868 individuals belonging to 30 species of breeding birds (checklist and data on abundances and on their frequencies at species level are reported in the Supplement Sect.S3).
Mean abundance and mean richness resulted in significant differences among the four age classes for all the forest habitats (Friedman test; Table 2), with highest values in class 4 at all sites.Analogously, Shannon-Wiener index also increased progressively from the youngest (1) to oldest (4) age classes.
Comparing mean abundance and mean richness between paired forest habitats for each age class, we observed significant differences (p < 0.05) or p values tending to significance (p < 0.09, Wilcoxon test) only at the higher age classes (3 and 4; Table 3).Comparing mean abundance and mean richness between contiguous age classes for different forest habitats, we observed significant differences at all sites (p < 0.05, Wilcoxon test; Table 4).
Regarding intra-habitat β w turnover at wood patch level, the younger age classes of coppice (1 and 2) showed the highest values (i.e. the younger the age class, the more locally    heterogeneous the habitat) compared to the oldest stages.We observed differences in pattern among forest habitats: indeed, in contrast to Turkey oak woods and beech forests (where β w species turnover was the highest in the age class 1), in chestnuts we showed an increase in species turnover in the forest renewal stage (age class 2; Table 2).
When we compared forest habitats, controlling for age class, we observed that mean abundance and richness significantly differ only in the most mature classes, while no difference occurs between recently cut woods (age classes 1 and 2).In the youngest age classes, forest habitats show a simplified structure which appears suitable only for a restricted number of species, which are more ecologically linked to shrubs, edge and open habitats and not specialised toward different tree species.Quantitative differences between forest habitats become significant only when these habitats markedly change their structure (oldest age classes: 3 and 4); i.e. when the woods grow older, the structural differences increase among different habitats, making new and different resources and niches available and favouring a quantitative differentiation in abundance and richness of bird communities.
Specifically, at species level, some hole-nesting species increases in abundance as tree size progressively increases, for example tits (Paridae), woodpeckers (Picidae), Certhia brachydactyla and Sitta europaea.On the contrary, some edge species inhabiting Mediterranean scrublands or coppices show an opposite trend, decreasing in abundance or disappearing with the increased forest age and structural complexity (e.g.Sylvia melanocephala, S. cantillans, Chloris chloris, Serinus serinus).Two species represent particular cases: Fringilla coelebs occurs only in the first and fourth ages of a forest, in both cases when tree density is low; conversely, Erithacus rubecula appears favoured by tree density, so its highest abundance values have been found in the intermediate ages.
Coppice management practices induce an increase in environmental heterogeneity at landscape and patch scale (Turner, 1989;McGarigal and McComb, 1995), affecting forest bird communities (Camprodon and Brotons, 2006;Castro et al., 2009;Dickson et al., 1993;Paillet et al., 2010).In our study, the younger age classes of coppice (1 and 2) also showed the highest level of intra-habitat β w turnover at wood patch level: therefore, the younger the age class, the more heterogeneous the habitat (hosting open areas, shrubs, isolated trees), in contrast to the oldest stages, where only mature trees with a closed canopy are present (see Webb et al., 1977;Crawford et al., 1981;McComb et al., 1989).However, we observed a different pattern among forest habitats.Indeed, in contrast to Turkey oak woods and beech forests (where β w species turnover was the highest in the youngest age class 1), in chestnuts we showed an increase in species turnover in the forest renewal stage (age class 2) due to different coppice management practices (Spinelli et al., 2010).Indeed, the physical structure of commercial coppiced chestnuts is substantially different from that of other types of broadleaved coppiced woods (Fuller and Moreton, 1987): in these woods, during the renewal stage there is rapid growth of the shrub layer with large leaves, increasing habitat heterogeneity and favouring many species linked to these new shady conditions (e.g.Turdus merula, Erithacus rubecula, Troglodytes troglodytes), with a consequent increase in species turnover at patch scale and at community level.
The bird communities associated with forest habitats are subject to different effects of forestry, depending on their level of specialisation (Dickson et al., 1993).Our communities were mainly composed of forest species.Therefore, patterns observed for overall bird assemblages have also been confirmed for the guild of strictly forest species: their structure significantly changed in all forest types, both in terms of mean abundance and richness along age groups, with a gradually increasing trend toward the older age class.
Forest species responses to structural changes were also evident.Forest species can be divided into two groups, (i) forest species sensu lato, i.e. generalist species linked to forest, wooded mosaics and wood-edge habitats, and (ii) forest interior species, i.e. specialised species that nest only within the interior of the forest and tend to avoid edge habitats (Whitcomb et al., 1981;Villard, 1998).The generalist forest species, such as tits (Paridae), Regulus ignicapilla, Certhia brachydactyla, Troglodytes troglodytes and Columba palumbus, are already present in the initial forest age (1 and 2) with very low abundance values, but grow with increasing forest age and structural complexity.A different case is the forest interior (sensu stricto Villard, 1998); for example woodpeckers (Picidae) and Sitta europaea are completely absent in the initial forest ages, appearing only in the fourth forest age in the case of coppice (chestnut and Turkey oak) and also in the intermediate forest ages in the high forest (beech).This is because in the intermediate forest ages of beech, woods aged over 30-50 years and they have characteristics suitable for these species.These species are known as interior hole-nesting birds and are highly sensitive to tree structure, forest changes and related disturbances (e.g.Mc-Collin, 1993;Mikusinski and Angelstam, 1997;Matthysen, 1998;Pasinelli, 2007).
Our study analysed communities almost entirely composed of forest species and found an inverse pattern to that reported by Fuller and Moreton (1987) and Fuller and Henderson (1992), in which total abundance and diversity were highest in the youngest coppiced woods.Independently of tree species composition, older age class forests showed forest bird communities characterised by the highest values of diversity metrics (abundance, richness and Shannon-Wiener diversity), which were highly sensitive and specialised for more complex vegetation structure.
To conclude, our work shows that in forest bird communities when increasing age of the forests (i) habitat heterogeneity (β w turnover) decreased, (ii) all the diversity metrics increased, and (iii) differences in mean abundance and richness between different forest habitats become significant.Moreover, chestnuts showed a peculiar trend due to its characteristic pattern in tree growth.
However, these patterns could be dependent on forest types, on geographic and climatic conditions and contexts (Mediterranean area), on peculiar species composition (also due to distributional factors of species at regional scale) and on the local history of coppicing.Moreover, although our stratified sampling design allowed consistent data patterns to be obtained on different treatments (multiple age classes for each forest habitat), it is probable that increasing the number of sampling points for each treatment could add information on some rare species not recorded in our analysis.Therefore, long-term studies at wider spatial scales will be desirable to corroborate our patterns, by providing support to select sensitive bird species in managed forest landscapes (see Villard and Jonnson, 2009).

Figure 1 .
Figure 1.Map of the study area, depicting Italy (top right), the Lazio region and the sampling area on the hills north-west of Lake Bracciano (bottom left, dotted).

Figure 2 .
Figure 2. Mean abundance of forest bird species (see Table 2) in the different forest habitats (age class and tree composition).Chestnut in white, Turkey oak in grey and beech in black.Error bars represent the standard deviation.
Figure 2. Mean abundance of forest bird species (see Table 2) in the different forest habitats (age class and tree composition).Chestnut in white, Turkey oak in grey and beech in black.Error bars represent the standard deviation.

Figure 3 .
Figure 3. Mean richness of forest bird species (see Table 2) in the different forest habitats (age class and tree composition).Chestnut in white, Turkey oak in grey and beech in black.Error bars represent the standard deviation.
Figure 3. Mean richness of forest bird species (see Table 2) in the different forest habitats (age class and tree composition).Chestnut in white, Turkey oak in grey and beech in black.Error bars represent the standard deviation.

Table 1 .
Structural traits of forest habitats in the study area (see Methods for details).DBH is diameter at breast height.

Table 2 .
Structural parameters of the bird community in the different forest habitats (Ch is chestnut; Oa is Turkey oak; Be is beech woods).S is species richness; S mean is mean species richness (and standard deviation, SD); N is total abundance; N mean is mean abundance (and SD); % FrFor is proportion (in percentage) of forest species frequency; H is Shannon-Wiener diversity index; β w is intra-habitat β diversity.Coefficient (χ 2 ) of Friedman test and p values are reported.Significant values are in bold.

Table 3 .
Comparison of mean abundance and mean richness between paired forest habitats (Ch is chestnut; Oa is Turkey oak; Be is beech woods) for each age class (from 1 to 4).Z and p values (Wilcoxon test) are reported.Significant values are in bold.*p < 0.01

Table 4 .
Comparisons between mean abundance and mean richness between contiguous age classes (from 1 to 4) for different forest habitats.Z and p values (Wilcoxon test) are reported.Significant values are in bold.