DIVERSITY AND ECOLOGICAL DIFFERENTIATION OF OAK FORESTS IN NW THRACE ( TURKEY )

Quercus robur, Q. frainetto, Q. cerris and Q. petraea that predominate in forests in the Thrace region, a bridge between the Balkans and Anatolia, were sampled, elaborated and classified. The ecological conditions were estimated by bio-indicator values. Oak forests can be divided into four groups: Q. robur-Fraxinus angustifolia forests thriving in floodplains, Q. petraea forests found at higher altitudes, Q. frainetto-Carpinus orientalis forests appearing in the warmest and driest sites and Q. frainetto forests on more humid sites. It was established that the most important topographic factor is altitude, while slope and aspect are of minor importance.

In Thrace, oak-dominated forests appear in floodplains, in the lowlands and hilly regions, while in the mountainous zone they are in contact with Fagus orientalis forests (Yarci, 2000).Floodplain forests are specific forest communities with a particular species composition related to the habitat formed on an alluvial plain.There are some regional differences, but the communities have an azonal character and, in spite of some regional characteristics, are individually fairly unique all over the Mediterranean region (Brullo and Spampinato, 1999;Klimo and Hager, 2001).The vegetation in the lowland and hilly region of Thrace is dominated by mixed oak forests called Salvio forskhali-Quercetum cerris quercetosum frainetto (Akman et al., 1979;Quezel et al., 1992;Aydin et al., 2008) that resemble the oak forests of the Balkan Peninsula (Bergmeier and Dimopoulos, 2008;Butorac et al., 2008;Čarni et al., 2009).
Apart from its historical importance, this region is important from a phytogeographical point of view, since the "oriental" component of the Balkan flora reached southeast Europe via the Thracian plain while some species only reached Thrace and did not pass further to the west (Magyari et al., 2008).It was therefore expected that the vegetation would mediate between the vegetation of Anatolia and that of the Balkans.
There are only a few publications about the forest vegetation of Thrace (Yaltirik et al., 1983;Yarci, 2000); some additional information is also available from surveys of the forest vegetation of Turkey (e.g.Mayer and Aksoy, 1986;Akman, 1995;Çolak et al., 2007), but no work exists that offers an insight into the floristic composition and ecological conditions of the oak forests of Thrace.Such a work is needed to enlarge the general knowledge of the flora and vegetation of southeast Europe, especially in the transitional zone to Anatolia.Knowledge of forests, their species composition and ecological conditions would also offer a basis for forestry practice and maintaining of biodiversity.
The aim of our work was to sample and elaborate the oak forests in NW Thrace.We tried to discover which oak species make up the forests in the region, to define their floristic and ecological diversity and to establish their distribution pattern.We attempted to identify the most important ecological and topographical factors that cause the diversity of oak forests.We treated the forests in the broader geographical context of the transitional zone between southeast Europe and Anatolia.The research provides a basis for nature protection and for the sustainable management of forests.

Study area
The study was undertaken in NW Thrace, which is composed of a mountainous part and lowland.The Istranca (Yildiz) mountain range is parallel to the coastline and is covered in the lower part by mixed oak forests.In the lowland, there is the well-known Igneada region, one of the most important plant areas of Turkey, due to its biological and ecological richness.It contains floodplain forests, lakes, swamps, scrub communities and sand dunes.However, the upper part of the mountains is mainly dominated by Fagus orientalis.(Özhatay et al., 2003) The average annual rainfall is about 800 mm and the average temperature is 13 o C. The hottest month is August and the coldest month is February (Anon., 2006).According to the Thornthwaite (1948) climate system, the research area has a humid and mesothermal sea climate.The bedrock is mainly formed by sedimentary rocks such as alluvial, calcareous rocks, noncalcareous and pliocene sedimentary rocks, dunes and siltstone flischists (Kantarcı, 1979;Sevgi, 2005).

Analysis of vegetation data
The field work was carried out between 2003 and 2006.We sampled thermophilous oak forests dominated by Q. frainetto, Q. petraea and Q. cerris, as well as floodplain forests with a high abundance of Q. robur.We tried to include in our analysis all types of oak forests in the region.We chose homogenous sampling plots with an area of 400 m 2 .The protocol of each plot includes general, topographic and other data of individual plots, such as altitude, inclination, aspect, vegetation cover (total and of individual layers) and a list of all vascular plants, in which a cover value was assigned to each species according to the nine degree Braun-Blanquet scale (Braun-Blanquet, 1964;Westhoff and van der Maarel, 1973).
The samples (hereinafter relevés) were stored in the TURBOVEG database management program (Hennekens and Schaminée, 2001).The hierarchical classification of the data set was carried out in the computer program SYN-TAX (Podani, 2001).The Bray-Curtis distance was used as a resemblance measure for the analysis and the betaflexible algorithm with β:-0.25 for dendrogram construction.Different levels of division were accepted in the dendrogram, resulting in four clusters interpretable in terms of ecology.Additionally, the diagnostic species of the accepted clusters were identified by a fidelity measure in the JUICE program (Tichý, 2002).The threshold of the phi value was subjectively selected at 0.50 for a species to be considered as diagnostic (Chytrý et al., 2002).Determination of the higher syntax was performed according to Aydin et al., (2008) and the characteristic species of the higher syntax were defined in accordance with Akman (1995).
The results of the classification were visualized by ordination techniques in the CANOCO 4.5 package (ter Braak and Šmilauer, 2002).Principal Component Analysis (PCA), which is an indirect ordination method assuming a linear response of species to the environment, was run due to the low heterogeneity in the matrix of species (Lepš and Šmilauer, 2003).Since it was seen in the first step of the analysis that the floodplain Q. robur-Fraxinus angustifolia forests form a strictly separated group, distinct from other thermophilous oak forests, they were excluded from further analysis.
PCA analysis was then performed again only for thermophilous oak forests, in order to reveal their ordination more clearly.Unweighted average indicator values were used for further interpretation of ecological conditions (Zelnik and Čarni, 2008).Values were calculated using bioindicator values (Pignatti, 2005) in the Juice program.These and topographic factors, such as aspect, altitude and inclination, were used as explanatory variables and projected passively onto the ordination plane.Although the bio-indicator values were selected for Italy, they are useful and often used also for Balkan vegetation (Košir et al., 2008;Tsiripidis et al., 2007).Correlations between PCA relevé scores and explanatory variables were calculated using the non-parametric Kendall coefficient in STATISTICA (Anon., 2007).
Additionally, redundancy analysis (RDA), a linear and direct ordination technique used because the gradient of relevés is narrow (Lepš and Šmilauer, 2003), was chosen to show correlations between relevés and variables that were measured in the field (aspect, altitude and inclination).The unrestricted Monte Carlo test with 9999 permutations was also used to test the statistical significance of the variables.
Identification of the species was made according to the Flora of Turkey (Davis, 1965(Davis, -1985;;Davis et al., 1988) and a new syntaxon was described in accordance with the International Code of Phytosociological Nomenclature (Weber et al., 2000).

Classification
The classification of the database shows four clusters (Fig. 1).Diagnostic species of these clusters and vegetation table of the relevés are shown in Table 1.
Cluster 1 includes Q. robur-Fraxinus angustifolia subsp.oxycarpa-dominated forests.It is the most distinctly separated community in the dataset, characterized by many species, such as Ulmus minor, Carpinus betulus, Rumex conglomeratus, Carex remota and many others that show the high moisture status of the site.
Cluster 3 contains Q. frainetto-Carpinus orientalis-dominated forests.The diagnostic species of this group is Carpinus orientalis, which appears on the warmest and driest sites within the studied forest types.
Cluster 4 includes Q. frainetto-dominated forests.Diagnostic species include heliophilous herb species, such as Galium verum, Brachypodium pinnatum and Trisetum flavescens, which indicate the favorable light and moisture conditions.Fig. 2 shows that the floodplain Q. robur-Fraxinus angustifolia-dominated forests have a very different floristic composition from the thermophilous oak forests, due to their hygrophilous character.They were therefore excluded from further analysis.
The group of thermophilous oak forests was again elaborated using PCA with passive projection of explanatory ecological factors, in order to un-derstand better the ecological conditions (Fig. 3a).Kendal's correlation coefficient revealed that ecological and topographic factors have a significant effect on thermophilous oak communities along both axes (Table 2).The first axis is significantly correlated with ecological factors, such as temperature, nutrients and moisture, whereas the second axis is related to topographic aspects, such as altitude, aspect, inclination and also light.
In the diagram (Fig. 3a), the left part shows Q. petraea-dominated forests which are found at higher altitudes and the ecological conditions are least propitious, since moisture and nutrients are not available.In Q. frainetto-dominated forests, conditions are most favorable.There is the highest light, moisture and nutrient availability.Q. frainetto-Carpinus orientalis-dominated forests are found in the warmest sites with a well pronounced continental effect.
In the species ordination diagram (Fig. 3b), species of drier and cooler Q. petraea-dominated forests are on the left.These are followed in a clockwise direction by forests found in more favorable ecological conditions, dominated by Q. frainetto.Finally, species of the most thermophilous and continental forests, dominated by Q. frainetto-Carpinus orientalis, are shown.

Table 1.
Vegetation table of the relevés .Group numbers correspond to the community numbers mentioned in the Figure 1.Legend: upper tree layer (1), lower tree layer (3), scrub layer (4) and herb layer (6).Coverage of the layers and localities of the relevés are submitted in the Appendix.

Ordination
An attempt was also made to evaluate the importance topographical factors measured in the field, in order to understand the distribution of samples in the ordination diagram (Table 3).As a result of this analysis, it was seen that altitude is the only significant topographical factor, and aspect and inclination are of minor importance.

DISCUSSION
The distribution pattern of oak forests in the region was ascertained.Q. robur-Fraxinus angustifoliadominated forests can be found on nutrient rich, humid sites in floodplain areas.They are characterized by species with a Euro-Siberian distribution pattern.They are very different from other oak forests in the region.Q. petraea-dominated forests can be found on the coolest, least favorable and highest sites of thermophilous oak forests in the region.They correspond to the physiological optimum of Q. petraea, which is less adapted to warm, dry habitats than Q. frainetto (Manes et al., 2006;Sisó et al., 2001).Q. frainetto-Carpinus orientalis-dominated forests appear on the warmest sites in the region with moderate inclination.These forests are the most continental ones here, as is evident from the presence of Carpinus orientalis, which is a species of continental character (Blasi et al., 2001).Q. frainetto-dominated forests are found on the most favorable conditions of all thermophilous oak forests.The high light availability may be related to the high degradation of these forests.
Q. robur-Fraxinus angustifolia-dominated forests are found on an alluvial plain.This vegetation is to some extent azonal and its composition is therefore similar to that in other south European regions (Brullo andSpampinato, 1997, 1999;Pavlov and Dimitrov, 2002;Vukelić and Baričević, 2004).We classified the communities into Carici remotae-Fraxinetum oxycarpae, which is an association of the alliance Alno-Quercion, with the synonym Fraxinion angustifoliae.
Thermophilous oak forests in the region show a typical feature of the oak forests of southeastern Europe.Thermophilous oak forests in the central Balkans are divided into two groups/alliances, one of low altitudes dominated by Q. frainetto and Q. cerris (alliance Quercion frainetto), and the other dominated by Q. petraea and Q. cerris (alliance Quercion petraea-cerris) at higher altitudes (Kojić et al., 1998;Čarni et al., 2009).In our region, the division into two alliances is not suitable and we classified all thermophilous forests into a single alliance Quercion frainetto.Barbero et al., (1976) divided Q. frainetto-dominated forests in the Balkans into two groups/alliances, of which one is more continental (alliance Quercion frainetto) and the other Mediterranean (alliance Melitti-Quercion frainetto).Q. frainetto-dominated forests of the European part of Turkey should be classified into the continental group (Akman et al., 1979).Thermophilous oak forests in the region show a typical feature of the oak forests of southeastern Europe.Thermophilous oak forests in the central Balkans are divided into two groups/alliances, one of low altitudes dominated by Q. frainetto and Q. cerris   (alliance Quercion frainetto), and the other dominated by Q. petraea and Q. cerris (alliance Quercion petraea-cerris) at higher altitudes (Kojić et al., 1998;Čarni et al., 2009).In our region, the division into two alliances is not suitable and we classified all thermophilous forests into a single alliance Quercion frainetto.Barbero et al., (1976) divided Q. frainetto-dominated forests in the Balkans into two groups/alliances, of which one is more continental (alliance Quercion frainetto) and the other Mediterranean (alliance Melitti-Quercion frainetto).Q. frainetto-dominated forests of the European part of Turkey should be classified into the continental group (Akman et al., 1979).
Q. frainetto and Q. frainetto-Carpinus orientalis-dominated forests can be classified within the sub-association of Salvio forskhali-Quercetum cerridis quercetosum frainetto.This community has been described in the eastern Marmara region (Akman et al., 1979) but it seems that it is distributed over the majority of the Marmara Region.It shows common features with the forests of the Balkans, with many species characteristic of thermophilous deciduous forests.However, it is differentiated by some Anatolian species, such as Salvia forskhali, which are not distributed in other parts of the Balkans.
On the basis our analyses, we divided Salvio forskhali-Quercetum cerridis quercetosum frainetto into two ecological variants: Carpinus orientalisvariant (Q.frainetto-Carpinus orientalis-dominated forests) and Galium verum-variant (Q.frainettodominated forests).The Carpinus orientalis variant is found on the driest and most continental sites and is characterized by co-dominance of Carpinus orientalis.The Galium verum variant is found in the most favorable site conditions and is characterized by the presence of G. verum, Trisetum flavescens and Brachypodium pinnatum.Q. petraea subp.iberica-dominated forests show similar features to those from Belgrade Forest near Istanbul, which was incorrectly classified as Q. petraea subsp.iberica-Lathyrus niger association (Yaltirik et al., 1983;Weber al., 2000).Aydin et al., (2008) classified Q. petraea s. iberica-dominated forests from the same locality as Salvio forskhali-Quercetum cerridis quercetosum frainetto.According to our analysis, Q. petraea subsp.iberica-dominated forests show a different floristic composition and ecological conditions from Salvio forskhali-Quercetum cerridis quercetosum frainetto and should therefore be treated as an independent association.Since Lathyro nigri-Quercetum petraea (Baričević et al., 2006) already exists, it was decided to classify the Q. petraea subsp.iberica-dominated forests into Taneceto cinerei-Quercetum petraea subsp.ibericae ass.nova.These communities have a different floristic structure from the Q. petraea dominated forests described in Anatolia (Yurdakulol et al., 2002;Türe et al., 2005).The holotype of this association is relevé 17 in  Akman et al., ex Quezel et al., 1992 var. Carpinus orientalis 4) Salvio forskhali-Quercetum cerridis quercetosum frainetto Akman et al., ex Quezel et al., 1992 var. Galium verum In conclusion, this work brings new data about oak-dominated forests in Thrace, which have not previously been elaborated.This enlarges our knowledge of the forest vegetation of Thrace and provides a basis for nature protection and for the sustainable management of the forests.At the same time, it also confirmed that the standard Braun-Blanquet method is a useful tool for forest site classification in Turkey.

Figure 3b .
Figure 3b.PCA ordination of species of thermophilous oak forests.t1, t2, s, and h symbolize upper tree, lower tree, scrub and herb layers, respectively; herb species always appear in the herb layer.

Table 3 .
Results of forward selection: environmental variables and the percentage of total variance of species data explained by RDA analysis.