Diversity, ecological structure, and conservation of the landbird community of Dadia reserve, Greece

Study area

The study area is situated in north-eastern Greece (40°59′−41°15′ N, 26°00′−26°19′ E). It covers 430 km², almost all of which belongs to the Dadia Nature Reserve. It is a hilly area with elevation ranging from 10 to 650 m. The climate is submediterranean. Temperature ranges from −19° to 40 °C and the mean annual rainfall varies between 556 mm and 916 mm (Adamakopoulos et al., 1995). Forests are dominated by Aegean pine (Pinus brutia) and black pine (Pinus nigra), oaks or a mixture cover 75% of the reserve, followed by agricultural land (16%), shrubs and mosaics (5%), and open areas consisting of grasslands and heaths (4%). The reserve was declared in 1980 and consists of two strictly protected zones (72.9 km²), which, in combination, have 85% cover of pinewoods and mixed pine–oak woods and are of great importance as nesting sites for Eurasian black vultures (Poirazidis et al., 2004). The buffer zone of the reserve (351.7 km²) has 72% forest cover and has more open or semi-open areas. A detailed map of the vegetation types in the study area is presented in Kati et al. (2004c).

Sampling

We distinguished 21 vegetation types in the study area, using the Corine typology database (Devillers & Devillers-Terschuren, 1996) as a reference system (19 Corine habitat types present in the study area). In two cases (Corine codes 41.733 and 4285A) we further distinguished two vegetation subcategories, on the basis of vegetation physiognomy (bushy undergrowth or not). One site or usually two sites (indicated with ‘a’ or ‘b’ next to the site code) were randomly selected to represent each one of the 21 vegetation types of the study area. We sampled 36 sites with 310 point counts (Blondel et al., 1970; Bibby et al., 1992). In a point count, the territorial-breeding song of a male represents a future breeding pair and is therefore counted as two individuals (Bibby et al., 1992). Every call different from the male breeding song is counted as one individual. We recorded all bird species seen or heard within a circle of 100-m radius. Each count lasted for 10 min and took place between 30 min before sunrise and 4 h after sunrise. In order to sample both sedentary and late migrant species, we conducted two sets of 155 point counts between 15 April to 15 May and 15 May to 15 June, 1999. Because of the differences in area extent sampled, we could not have the same sampling effort at all sites and adjusted the sampling effort based on the site's area. We sampled 25 sites of 20 ha with five point counts, and 11 smaller sites with two to four point counts (Table 2). Sampling stations were at least 200 m away from each other and 100 m away from the site edge.

Each point count was located at the centre of a quadrat of 50 × 50 m, within which we estimated the overall vegetation cover of trees and bushes (> 0.5 m), in order to describe the general vegetation physiognomy of the site sampled (Table 1). The vegetation cover is defined as the total vertical projection of the crown area of trees and bushes, using the Braun–Blanquet scale for cover estimation: 1 = 1–5%, 2 = 6–25%, 3 = 26–50%, 4 = 51–75%, and 5 = 76–100%. Based on Kent & Coker (1994), we characterized the four main vegetation layers as follows: (1) trees > 7 m, (2) low trees (4–7 m), (3) high bushes (2–4 m), and (4) bushes (0.5–2 m). We defined the dominant vegetation layers as the ones contributing more than 25% to the overall vegetation cover (Table 1).

Data analysis

The species diversity of sites was estimated in terms of species richness (S), weighted species richness (WS), and Shannon–Weiner index (H) (Magurran, 2004). Weighted species richness is the species richness of the site, with each species having a different weight based on its conservation status (SPEC category) (BirdLife International, 2004). We gave a standard weight (w = 1) to the species of SPEC 4 category, which are concentrated in Europe and have either a favourable conservation status or are not covered by any SPEC category. We gave double weight (w = 2) to the species of SPEC 3 category, which are not concentrated in Europe but have an unfavourable conservation status. Finally we gave quadruple weight (w = 4) to the species of SPEC 2 category, which are concentrated in Europe and have an un-favourable conservation status and to the species of Annex I of the European Directive 79/409 EU, which includes all bird species subject to special conservation measures in Europe. We calculated H of each site using as relative abundance of bird species the maximum abundance of the two values recorded in the two sampling seasons.

Using the r3 software package (Legendre & Vaudor, 1991), we analysed the bird community composition by creating a similarity matrix of the samples on the basis of relative abundance data (Steinhaus coefficient of similarity — S17). A distance matrix was produced from the similarity matrix, and it was used as an input into the spatial visualization in the Euclidian space, using Principal Coordinate Analysis with corrected eigenvalues (DistPCoA), a method that resolves the problem of ‘non-Euclideanarity’ (Legendre & Anderson, 1998). We used the samples’ coordinates of DistPCoA as an input into the k-means partitioning procedure, using the fastclus procedure (SAS, 1985). We applied the above procedure for k algorithm computation, k being the final number of clusters produced. Computation was repeated 1000 times, randomly reallocating samples as initial seeds at each run. From the n dimensions of samples’ coordinates we used the first ordination axes capturing a standard percentage of 85% of data variance, avoiding thus the noise of the last axes. The final output was the non-hierarchical up-to-down robust dendrogram with k clusters, including only samples with a level of persistence > 75%. These k clusters are also shown as 80% confidence ellipses in the ordination of point counts (SAS, 1985). Finally, we used the indicator value procedure (Dufrêne & Legendre, 1997) in order to find out the typical species that characterize each of the clusters. The indicator value of each species for a given cluster is calculated as: IndVal = A × B × 100, where A = mean number of the individuals across the sites of the cluster, which is the sum of mean number of the individuals in all clusters, and B = number of sites in the cluster where the species is present, which is the total number of sites in that cluster.

IndVal is a percentage that ranges between 0 and 100 and takes its maximum value when the species is present only in one cluster and in all sites of this cluster. All calculations were carried out using indval software (Dufrêne, 1999). A species is considered to be a ‘symmetrical indicator’ (IndVal > 50%) for one cluster when it is present in > 70% of the sites of the cluster and when > 70% of its individuals occur in the cluster. A random reallocation procedure (1000 iterations) of sites among site groups was used to test the significance level of IndVal (alpha = 0.05). In order to represent all the bird species in the study area, we drew a complementary network of sites starting with the most important site in terms of conservation and, using a stepwise procedure, added the sites that contributed new species to the network, giving priority to species with high conservation value (criteria WS and then S).

Diversity

The most important sites for bird conservation were highly heterogeneous sites, combining grassy openings, hedges, and wood plots, such as rural mosaic (A1) and mosaic sites (M1, M2), because they hosted the greatest number of species (S) and the greatest number of species of concern (WS) (Table 2). These mosaic character sites are proven also to be the most diverse, based on the Shannon's diversity index as a ranking criterion (M1a, A1b, A1a, M1b, M2).

To better compare the site species richness regardless of the variations in site area and the sampling effort, we also used the mean species richness of point counts as the main ranking criterion. Mosaic (M1a) and rural mosaics (A1a, A1b) again emerged as the three most important sites, followed by smaller sites such as grasslands with Phillyrea bushes (S3), serpentine grasslands (G3), or xeric grasslands (G2). These sites are openings within the expanded pine forest zone (Table 2). Oak matorral with shrubs (S2) also ranked as one of the most species-rich sites. Pinewoods hosted few species, including only two breeding species of conservation concern (Streptopelia turtur, Picus viridis), which also bred in the broadleaved woods of the study area.

Ecological structure

Considering all the 155 point counts together, the ordination procedure (DistPCoA) clustered the 10 point counts in field crops (A2), positioning them on the right part of the horizontal axis that explained 23% of the variation. Therefore, agricultural land with almost no tree or bush cover is clearly distinguished from the remaining sites. We then excluded these point counts and re-analysed our data set, in order to discern the environmental gradients within the remaining 145 point counts. Shaded sites with high vegetation cover were positioned on the left of the horizontal axis which explained 19% of the variation, semi-open sites in the centre, and open sites on the right. On the other hand, the vertical axis that explained 16% of the variation separated sites with tall trees from those with shrubs (Fig. 1). Hence, two major environmental gradients affect bird distribution in the study area: vegetation cover and vegetation height.

The first cluster set apart in the clustering procedure encompassed all point counts conducted in the field crops (A2). We recorded three bird species exclusively in the field crops: Melanocorypha calandra, Calandrella brachydactyla, and Anthus campestris (IndVal = 20%, 20% and 10%, respectively).

Figure 2 shows the main habitat types defined by their bird communities and the typical species for each cluster, using the 145 point counts analysed with the k-means and IndVal procedures. Some bird species are generalists and they are common throughout the study area, having no indicator value for a precise vegetation type (i.e. Fringilla coelebs, Turdus merula, Luscinia megarynchos, Erithacus rubecula, Garrulus glandarius, Oriolus oriolus, Parus major, Carduelis chloris, S. turtur, and Upupa epops). Combining the results from ordination and clustering procedures (1, 2), seven different bird habitat types can be distinguished: heaths, shrubs and low trees, pinewoods, broadleaved woods, mosaic sites and forest openings, rural mosaics, and poplar vegetation. Lullula arborea characterizes heaths, while Certhia brachydactyla is a typical species of the broadleaved woods. Emberiza cirlus characterizes the mosaic sites and the grassy openings inside forest areas. Five species are typical of agricultural fields separated by hedges and trees: Emberiza melanocephala, Galerida cristata, Hippolais pallida, Milaria calandra, and Sylvia communis. Emberiza hortulana characterizes areas with low trees, shrubs, and heaths, while Parus caeruleus characterizes forest habitats. No characteristic species exist for shrubs and low trees, pinewoods, and poplar habitats.

Species representation

All bird species found in the study area can be represented in a network of 10 sites. This network includes one site from each bird habitat type (Fig. 2), with the exception of rural mosaics that contribute two sites in the network. The network sites are prioritized as follows, with complementary weighted species richness of each site and complementary species richness (number of new species contributed to the network) given in parentheses: rural mosaic A1b (65, 35), mixed pine–oak wood F5b (21, 9), field crops A2b (17, 6), poplar tree line F10a (11, 8), xeric grassland G1b (5, 4), heath Ha (5, 4), high maquis S1a (4, 1), rural mosaic A1a (3, 2), beech wood F1a or F1b (2, 2), and mosaic M1b (2, 2).

Ecological structure

Vegetation cover and height determine the composition of the avifauna in our study area, which is in agreement with other studies in the Mediterranean region (Blondel et al., 1970; Prodon & Lebreton, 1981; Catsadorakis, 1997). The height gradient mainly corresponds to vegetation successional stage and the vegetation cover to disturbance intensity, caused either by humans (agriculture, logging, livestock grazing) or by natural processes (fire, native herbivore grazing).

Birds discriminate landscape features at a coarse scale, perceiving the 21 vegetation types of the study area as eight distinct bird habitat types: field crops, rural mosaics, semi-open mosaics and grasslands, poplar vegetation, broadleaved woods, pinewoods, shrubs, and heaths. The habitat typology as revealed by the landbird community does not necessarily correspond to the predefined vegetation classifications (Fig. 2), thus emphasizing the importance of planning habitat conservation from the perspective of the taxon of interest. Each of these bird habitat types is also represented in the complementary network constructed to include all bird species, with rural mosaics being represented twice. Kati et al. (2004b) also found that for various taxonomic groups, including landbirds, any number of sites selected at random with the condition that each site belonged to a different cluster, conserves more species than an equal number of sites selected at random from different vegetation types defined after Corine typology. Hence, bird typology provides us with a guideline for specific habitats to be maintained for avifauna conservation.

Even though some of the species recorded are generalists, there are 10 specialist species that are highly characteristic and strongly dependent on the habitat types they are found in, as they are found in almost all sites of that habitat type and rarely in others. Monitoring the populations of such ‘indicator’ bird species may be a cost-effective and efficient way to monitor the overall landbird community and their habitats.

Conservation of landbird diversity

The rural mosaics were the most diverse and important sites for conserving the breeding bird fauna in our study area. Cultivated areas are generally known to play a fundamental role in maintaining breeding bird diversity in the Mediterranean region (e.g. Farina, 1997; Suarez-Seoane et al., 2002). Not only human-dominated habitats can support a substantial proportion of native bird diversity (Hughes et al., 2002; Sodhi et al., 2005) but these birds in return also provide key ecosystem services such as seed dispersal and pest control (Sekercioglu et al., 2004). However, agricultural intensification has led to a dramatic decline in farmland bird diversity in many European countries (Pain & Pienkowski, 1997; Chamberlain et al., 2000; Donald et al., 2001), although not all of them (Fox, 2004). In Greece, the policy followed by the Ministry of Agriculture during the last 50 years aimed at the unification of small agricultural proprieties into expanded agricultural land in order to permit the use of large machinery and the intensification of crop production. Small properties were exchanged through a procedure of land reallotment, simultaneously removing ‘living fences’ and tree and bush vegetation that comprised the natural borders of former small-scale properties. Our results show that rural mosaics, defined as small fields and pastures separated by natural vegetation of thick hedgerows and tree lines were twice as rich in bird species number than intensified crop monocultures. Removing hedges and lines of trees, even exotic conifer species (Pithon et al., 2005), can have negative effects on farmland birds. Our study provides additional evidence supporting the European common agricultural policy (CAP) towards the maintenance of the rural mosaic landscape as a habitat of great importance for farmland breeding birds in the Mediterranean region.

We found that semi-open mosaic sites and forest openings are also among the most species-rich sites in our study area and host species of conservation concern. Mosaics are heterogeneous sites combining patches of woodland, shrubs, and pastures in a small area, whereas openings in forest enhance landscape heterogeneity. Spatial heterogeneity is often the main factor that increases avian species diversity within habitat and landscape scales (Roth, 1976; Huston, 1994; Bohning-Gaese, 1997; Farina, 1997; Sekercioglu, 2002).

Forest openings are also important habitats for other taxonomic groups in the study area, such as insects (Grill & Cleary, 2003; Kati et al., 2004c) and several raptor species that use them for hunting (Adamakopoulos et al., 1995; Poirazidis et al., 2004). The forested zone includes fewer landbird species, but large areas of core habitat may be important to maintain viable populations of certain specialized forest-dwelling species. There is a need for research in the Mediterranean region to determine the threshold size of forest openings that will not harm forest-dependent bird species and hinder forest regeneration (Sekercioglu, 2002).

We found that all bird species sampled in the year 1999 can be represented in a complementary network of 10 sites. Reserve designers usually target to maximize protected biodiversity while minimizing reserve size (Cabeza & Moilanen, 2001). One of the best practices to do so is to pick up complementary areas with the maximum combined species richness (Pressey et al., 1993; Margules & Pressey, 2000), rather than species-rich areas, or areas representative of the vegetation types that are found at local (e.g. Kati et al., 2004b) or regional scales (e.g. Lombard, 1995; Howard et al., 1998). In the current study, our network of 10 selected sites is not intended as a proposal for a new subnetwork within the existing protected area of the Dadia Nature Reserve. These 10 sites by themselves are unlikely to maintain viable populations of the bird species of the reserve as units isolated from their landscape context. This network, however, highlights habitats that are important for local bird conservation. The network also confirms the importance of monitoring and conserving different bird habitat types represented in the clustering procedure, given that at least one site from the different clusters is represented in the network.

Interestingly, most species-rich habitats are located in the buffer zone, which is less forested, than in the strictly protected zone, which is covered mostly by pinewoods (85% cover). The high value of the buffer zone has also been shown for other biological groups studied in the reserve (e.g. Grill & Cleary, 2003; Kati et al., 2004c). The main conservation value of pinewoods is for the maintenance of the Eurasian black vulture population (Poirazidis et al., 2004) as well as for some forest-dwelling raptor species, rather than for the landbird community. It is encouraging that the management plan of the reserve (Adamakopoulos et al., 1995), which targets mainly the conservation of its birds of prey, is also compatible with the conservation of the landbird community in general. The management plan proposes to conserve rural mosaics with hedges and woodland patches and to maintain forest openings in the core area through livestock grazing, woodcutting, and the reintroduction of natural herbivore populations. Hence, there is no conservation conflict in the reserve as far as management practices are concerned.

Conservation proposals

The current study implies the importance of horizontal heterogeneity for bird conservation at the local and landscape scales, as shown by the high species richness of mosaic character sites and of openings in the forested zone, respectively. We also provide evidence against land reallotment and agricultural intensification and emphasize the importance, for breeding birds, of conserving rural mosaics in the Mediterranean landscape. Our findings also indicate that the knowledge of bird community structure should be integrated into conservation decision-making focused on landbird communities.

An ecological analysis of bird community structure resulted in the identification of eight distinct bird habitat types and 10 indicator bird species that are highly dependent on the habitat types they breed in. We propose the landbird community to be integrated as a monitoring parameter in the ongoing pilot monitoring project of the reserve and we provide a list of indicator bird species to monitor. Monitoring these indicator species is an efficient method to monitor the ecological state of the landbird community in the reserve and it gives a more direct insight into bird habitat quality, than surveying all vegetation types defined after standard habitat typologies such as Corine. Finally, reserve authorities should be aware of the importance of the buffer zone for local landbird diversity and consider the high conservation value of rural mosaics, mosaic character sites, and forest openings, especially when putting the reserve management plan into practice. Our conclusions have broader implications, both for the conservation of landscape heterogeneity in the European countryside and for the use of birds as indicator species worldwide.