Bird community patterns in sub – Mediterranean pastures : the effects of shrub cover and grazing intensity

Shrubs are widely considered a threat to grassland biodiversity. We investigated the effects of shrub cover and grazing intensity on bird communities in sub–Mediterranean pastures in Bulgaria. The point–count method was used on 80 plots in open (< 10% shrub cover) and shrubby (approx. 20% cover) pastures under either intensive or extensive management (grazing intensity) from 2008 to 2009. We recorded a total of 1,956 observations of birds from 53 species. Main environmental gradients accounting for the bird community pattern were related to vegetation succession and land productivity. Bird species richness was higher in shrubby pastures than in open sites, while no effect was found in respect to total bird abundance. Bird species diversity (i.e. H’ index) was highest in extensive shrubby pastures. Shrubland specialists were positively affected by shrub cover and extensive management of pastures while grassland and woodland specialists showed no significant response to these factors. We conclude that a small proportion of shrubs within pastures may be beneficial for farmland birds and sustainable management of pastures could be achieved by greater flexibility of national agri–environmental schemes.


Introduction
Semi-natural grasslands are among the high nature value farming systems of conservation concern as they are biodiversity-rich and provide agricultural benefits through stock grazing and haymaking (Henle et al., 2008).These habitats were created under traditional agricultural practices but currently, due to agricultural intensification or land abandonment, they have now become significantly reduced in area in northern European (Pärt & Söderström, 1999), western European (Tucker & Heath, 1994;Fuller et al., 1995) and eastern European countries (Meshinev et al., 2005).
In the European Union, semi-natural grasslands are under the regulation of the Common Agricultural Policy (CAP) which was adopted in 1957 to increase agricultural production by ensuring sufficient food for all inhabitants and a fair standard of living for farmers (Verhulst et al., 2004).Implementation of the CAP resulted in a polarization of production areas by stimulating land use intensification in some areas (Donald et al., 2002) and leading to abandonment of other, marginally profitable areas (Bignal, 1998).It was found that intensification or abandonment of land management can greatly reduce biodiversity by threatening the survival of many species adapted to the diversity of structures and resources of high nature value farmlands (Sirami et al., 2007;Kleijn et al., 2009;Nikolov, 2010).Grassland bird populations for instance, declined sharply due to agricultural intensification in Europe over the past half century (Gregory et al., 2004;Donald et al., 2006).On the other hand, abandonment of land management benefits vegetation succession through the development of woody vegetation, providing benefits to shrubland and woodland birds whilst negatively affecting open-habitat specialists (Preiss et al., 1997;Suárez-Seoane et al., 2002;Pons et al., 2003;Verhulst et al., 2004).As a result, the development of shrubby and woody vegetation was considered a potential threat to grassland biodiversity, and the CAP strongly advised removal of these habitat features as a management recommendation (Boccaccio et al., 2009).In many countries (e.g.France, Sweden, Greece and Bulgaria), this measure was not tested but applied directly in the national agri-environmental schemes (Lefranc, 1997;Pärt & Söderström, 1999;Söderström et al., 2001;Kati & Sekercioglu, 2006;Nikolov, 2010).Indeed, in northern and southeastern Europe small covers of woody vegetation (≤ 20%) were found to increase avian species richness and diversity by favouring some threatened species (Pärt & Söderström, 1999;Söderström et al., 2001;Nikolov, 2010).
The main objective of this study was to test the effects of shrub cover and grazing intensity on farmland birds in sub-Mediterranean pastures.The obtained results may serve as a basis for more sustainable and regionally-oriented pasture management aiming to maintain species rich, diverse bird communities.

Study area
The study area covers the territory of the Special Protected Area (SPA) Besaparski Hills (147.7 km 2 ) in southern Bulgaria (42° 7' N-24° 23' E; fig.1).The landscape represents sub-Mediterranean limestone hills with an average altitude of 350 m a.s.l.(ranging from 184 m a.s.l. to 536 m a.s.l.) (Demerdzhiev, 2007).Most of the area is covered by arable land (about 50% of the territory) and by dry grasslands with some shrub heath (about 33% of the territory) and the rest of the territory is covered by vineyards and orchards (6%), wetlands (3%), stone pits (3%), urban areas and roads (3%) and small forests (1%).Grasslands are not fertilized and most of them are used for pastures (mainly for sheep and cattle).

Study design
Based on a digital map of the area (Bulgarian Society for the Protection of Birds, unpublished data), a total of 80 point-count stations were equally distributed and located randomly within two categories of pastures (open pastures with up to 10% shrub cover and shrubby pastures with more than 10% shrub cover) with the restriction that any two adjacent point-count stations should be a minimum of 250 m apart (Ralph et al., 1995).All study plots with difficult accessibility to the field were replaced using a second random selection.As a result, an aggregation of study plots in the eastern part of the study area appeared, but as the study plots were equally distributed between the studied pastureland categories (25 vs. 25 study plots in shrubby and open pastures, respectively) within the area of aggregation, we assumed that our data were not biased by spatial autocorrelation effects.After a pilot visit to the study area, we found that 41 point-count stations were located within open pastures and 39 in shrubby pastures.Supplementary data on grazing intensity within the studied areas was collected from the local agricultural authorities and studied plots were classified according to their grazing regime as intensively grazed (0.8 AU ha -1 ; n = 30 study plots) and extensively grazed pastures (0.2 AU ha -1 ; n = 50 study plots).Finally, we used 31 study plots in open and extensive pastures, 10 in open and intensive pastures, 19 in shrubby and extensive pastures and 20 in shrubby and intensive pastures.
Fieldwork was carried out during the breeding seasons of 2008 and 2009.Birds were sampled twice per year (in May and June), in the mornings (6:00-10:00 a.m.), under appropriate weather conditions and by the same observer (D.D.).The point count method (Gibbons & Gregory, 2006) was applied, with a counting period of 5 min and a radial distance of 100 m.All birds seen or heard were recorded.Individuals simply flying over the point-count stations and not foraging in flight were excluded from the analysis (Batáry et al., 2007).To investigate how different ecological groups of birds respond to vegetation composition within pastures in respect to their habitat specialization we classified birds as grassland, shrubland or woodland specialists, following Iankov (2007).The conservation status of birds at the European and national levels was described following BirdLife International (2004) and Spassov (2007), respectively (appendix).
Data on habitat composition were collected within a radius of 50 m centred on each point-count station.The relative cover of rocks and stones, arable land and vegetation layers was estimated visually and recorded in percentages (%).The following vegetation layers were recognized in the field: (1) grass, consisted mainly of Medicago spp., Trifolium spp., Sideritis montana, Chrysopogon gryllus, Dichnthium ischaemum, Eryngium campestre and Stipa capillata; (2) shrubs, consisted of woody vegetation up to 2 m height and dominated mainly by Paliurus spina-christi, Rubus sp., Jasminum fruticans, Juniperus oxycedrus; and (3) trees, consisted of woody vegetation above 2 m height and dominated mainly by Quercus pubescens.Elevation was recorded using a Global Positioning Systems unit (Etrex Summit).Studied pasture cat-egories differed significantly only in their grass cover and shrub cover (table 1).

Data analyses
Bird data were square root transformed and habitat variables were arcsine transformed to approach normal distributions.For comparisons of environmental variables between open and shrubby pastures, t-test for unpaired samples was performed using STATIS-TICA version 7.0 software package (StatSoft, 2004).To analyze bird species richness and overall bird abundance we used the mean values of the maximum numbers of species and individuals recorded at each point-count station during both visits in both years.Bird species diversity was calculated using the Shannon-Wiener diversity index H'.
Relationships between bird species and habitat characteristics were determined by Canonical Correspondence Analysis (CCA) computed in CANOCO 4.5 software (Ter Braak & Smilauer, 2002).Length of  bird data gradient was checked by preliminary detrended correspondence analysis (DCA) and unimodal ordination was applied even though the gradient was relatively short (i.e.2.82 for the first canonical axis), because this model better explained data variability and because the length of the gradient was close to the range for which both linear and unimodal methods work well (Lepš & Šmilauer, 2003).The Monte-Carlo permutation test was used to assess the statistical significance of canonical axes (Lepš & Šmilauer, 2003).
The effects of shrub cover, grazing intensity and their interaction on birds at community and ecological group levels were analysed by General Linear Models (GLM) in STATISTICA version 7.0 software package (StatSoft, 2004).For each of the studied dependent variables (i.e.bird species richness, H' diversity index and abundance) separate GLM was conducted, where pasture categories (in respect of shrub cover and grazing intensity) were categorical factors and studied habitat characteristics (see table 1) were continuous predictors.In the GLM, a Tukey HSD post-hoc test was used to determine significant differences between groups (α = 0.05).

Habitat composition and bird community pattern
A total of 1,956 individuals from 53 species were recorded in the semi-natural grasslands of Besaparski Hills SPA (appendix).The main environmental gradient in the studied habitat was related to vegetation succession (representing the transition of open to shrubby pastures, fig.2) and was represented by the first CCA axis accounting for 15.6% of bird data variability (species-environment correlation = 0.743).All grassland specialists, excluding the woodlark Lullula arborea, showed a positive association with the open semi-natural grasslands (fig.2).Shrubland and woodland species displayed the opposite pattern, being associated with semi-natural grasslands with an increased cover of shrubs and trees.However, these species were more widely spread along the gradient.Apart from grassland species, the open pastures also sheltered aerial feeders (e.g.European bee-eater Merops apiaster and barn swallow Hirundo rustica) or birds that forage in open landscapes (e.g.Spanish sparrows Passer hispaniolensis).
The second environmental gradient was related to land productivity, (represented by land conversion: higher cover of arable lands at the one extremity of the gradient and the less productive rocky fields at the other extremity, fig.2) and was represented by the second CCA axis accounting for 8.1% of bird data variability (species-environment correlation = 0.736).Most birds associated with this gradient were shrubland and woodland species: some of them benefited from arable mosaics (e.g. common cuckoo Cuculus canorus, common starling Sturnus vulgaris and European roller Coracias garrulus), while others were associated mainly with the less productive grasslands (e.g.ortolan bunting Emberiza hortulana, blackcap Sylvia atricapilla and European greenfinch Carduelis chloris).

Effects of shrubby vegetation
Species richness was positively affected by the cover of shrubby vegetation while the effect of grazing intensity was not significant ( Grassland and woodland birds did not show significant response to the cover of shrubs within pastures or grazing intensity, while the low proportion of shrub cover and extensive grazing of pastures were found to increase the species richness and abundance of shrubland birds (table 3).
Regarding conservation status, open and shrubby pastures sheltered similar numbers but different species from conservation priority.Of the 15 species associated with open pastures (left sector of the biplot in fig.2), three are included in Annex 1 of the Directive on the conservation of wild birds (Directive 2009/147/EC), 10 have an unfavourable status in Europe and three are known to be decreasing in Bulgaria.Of the 17 species associated with shrubby pastures (right sector of the biplot in fig.2), five are included in Annex 1 of the Directive 2009/147/ EC, nine have an unfavourable status in Europe and three are known to be decreasing in Bulgaria.Fig. 2. Ordenación bidimensional por análisis canónico de correspondencias (CCA) que relaciona las abudancias de aves con las características del hábitat en los pastos submediterráneos de tierras bajas del sur de Bulgaria.Los dos primeros ejes canónicos responden del 23,9% de la variabilidad de los datos de las aves, y todos los ejes son estadísticamente significativos (test de Monte-Carlo, basado en 499 permutaciones al azar, F = 2,06, p = 0,02).Las variables ambientales se indican por medio de flechas.(Sólo se muestran las especies de aves que se ajustan > 5% al modelo; para los acrónimos de las especies de aves, ver el apéndice.)

Birds and shrubby vegetation cover
Our results demonstrate that a small proportion of shrubby vegetation (ca.20%) within semi-natural grasslands may increase the species richness of bird communities in the sub-Mediterranean pastures of southern Bulgaria.This finding is consistent with the results from dry pastures in northern Europe (Pärt & Söderström, 1999) and upland pastures in southeastern Europe (Nikolov, 2010) where retention of 10-15% shrub cover within pastures is advised as beneficial for farmland birds dependent on shrubs.This phenomenon could be explained by the increased habitat complexity within shrubby pastures, which provides more varied resources to bird species for nesting, searching for food, displaying (Verhulst et al., 2004) or escaping from predators (Shaefer & Vogel, 2000).This finding contributes to the concept that habitat heterogeneity is a key predictor for species richness within farmlands (Benton et al., 2003;Kati et al., 2009).Most of the positive effects of shrubby vegetation upon the structure of bird communities could be attributed to shrubland birds.Often this is a post-factum effect observed after land abandonment and the resulting secondary succession of the vegetation (Preiss et al., 1997;Suárez-Seoane et al., 2002;Verhulst et al., 2004; but see Batáry et al., 2007).However, a limited presence of shrubby vegetation (ca.20% cover) had no significant effects on grassland specialists.This was not expected as the presence of shrubs and trees within pastures reduces the overall area of the prime habitat for this group of birds.In western and central Europe, grassland bird abundance was observed to decrease as a consequence of the reduction of open grassland habitats (Preiss et al., 1997;Brotons et al., 2005) and it has also been found that increase in habitat heterogeneity may suppress the abundance of grassland specialists (Batáry et al., 2007).Possible ecological mechanisms explaining this pattern include an increasing predation risk for some grassland specialists due to the high vegetation cover in the surroundings (Shaefer & Vogel, 2000) or increased nest predation (Suárez & Manrique, 1992).In our study, the lack of negative effects of shrubby vegetation cover on grassland specialists may be explained by the relatively low cover of this habitat feature within studied pastures (about 20% mean cover for shrubby pastures; see table 1).

Conversion of pastures into arable land
Although several species (e.g. common cuckoo, common starling and European roller) were positively affected by the presence of arable lands, this should not be misinterpreted as a good reason for the conversion of grassland habitats into arable lands.In our study, the only threatened species associated with Table 2.The effects of grazing intensity (GI), shrubby cover (SC) and their interaction (INT) on bird community parameters in semi-natural grasslands of SPA Besaparski Hills (GLM, StatSoft, 2004): * Significant P-values are in bold; ** Between-group comparisons were determined by applying Tukey HSD post-hoc test, only significant differences (P < 0.05) are shown.
Table 2. Los efectos de la intensidad de pastoreo (GI), la cubierta arbustiva (SC) y su interacción (INT) con los parámetros de las comunidades de aves en las praderas seminaturales del área protegida especial de las colinas de Besaparski (modelos lineales generales, GLM, StatSoft, 2004) (Avilés & Parejo, 2004).Furthermore, some grassland specialists, including calandra lark Melanocorypha calandra and short-toed lark Calandrella brachydactyla, which are from high conservation priority within Natura 2000 network, are negatively affected by the presence of arable lands in grassland-dominated landscapes (Brotons et al., 2005).Therefore, our results suggest that it may be possible to support small parcels of arable land as a part of the rural mosaic within SPAs, but this practice should be adopted with caution and strictly controlled, as it is recognized as a major cause of the loss of the semi-natural grassland habitats (Robertson et al., 1990) and one of the main threats to the local avifauna (Demerdzhiev, 2007).

Bird conservation in sub-Mediterranean pastures
The CAP was implemented rapidly in many countries of the European Union, and most agri-environment schemes were applied without sufficient testing at national scales (Wrbka et al., 2008;Stoate et al., 2009;  (Tucker & Evans, 1997).Some of these effects on birds have been investigated (e.g.Batáry et al., 2007;Herzon et al., 2008;Nagy et al., 2009), but they are regionally dependent and should not be directly extrapolated to represent other locations with different cultural, economical and landscape characteristics (Nikolov, 2010).For instance, in many countries the removal of shrubby vegetation from pastures was promoted as an agricultural practice by CAP (Boccaccio et al., 2009), being considered as a threat to grassland biodiversity (Preiss et al., 1997;Stefanović et al., 2008).However, several studies from northern and south-eastern Europe have provided sound evidence that the availability of shrubby vegetation within semi-natural grasslands may be beneficial for the local avifauna (Pärt & Söderström, 1999;Söderström et al., 2001;Kati & Sekercioglu, 2006;Nikolov, 2010).Our results support this statement, demonstrating that from a conservation viewpoint open and shrubby pastures benefit a similar number of species of high conservation priority.We found that the small proportion of shrubby vegetation within pastures should not be considered as a threat, but as a potential factor increasing the conservation value of the protected area through the addition of some non-grassland threatened species to the existing typical grassland avifauna.Therefore, a possible way to counteract the negative effects of CAP on avian diversity in sub-Mediterranean areas could be at the level of national agri-environmental schemes (Verhulst et al., 2004) by providing higher flexibility of national standards at the regional scale (Wrbka et al., 2008).Particularly, regarding protected areas in the ecological networks this could be done by elaboration of management plans and zoning of agricultural activities at local scale.Finally, to ensure the effective long-term conservation of birds that are dependent on pastoral landscapes, it is crucial to assess the potential for resulting conflicts between intended outcomes for farmers and biodiversity (Henle et al., 2008).Therefore, we strongly advise further investigation into the stock holder's production losses in relation to the availability of shrubs within pastures, and the potential opportunities for the compensation of these losses.

Fig. 1 .
Fig. 1.Location of the study area (SPA Besaparski Hills) in Bulgaria and distribution of point-count stations within the semi-natural grasslands.

Fig. 2 .
Fig. 2. Two-dimensional ordination by CCA relating bird abundances to habitat characteristics in the sub-Mediterranean lowland pastures in southern Bulgaria.The first two canonical axes account for 23.9% of bird data variability and all axes are statistically significant (Monte-Carlo test based on 499 random permutations, F = 2.06, p = 0.02).Environmental variables are indicated by arrows.(Only brid species with fit > 5% in the model are shown; for bird species acronyms see the appendix.)
n = 10).Total bird abundance showed no significant response to shrub cover or grazing intensity (table 2).

Table 3 .
The effects of grazing intensity (GI), shrub cover (SC) and their interaction (INT) on species richness and abundance of ecological groups of birds in respect of their habitat specialization in pastures of Besaparski Hills, Bulgaria (GLM, StatSoft, 2004): * Significant P-values are in bold; ** Betweengroup comparisons were determined by applying Tukey HSD post-hoc test; only significant differences (P < 0.05) are shown.