SEEFOR 16(2): early view
Article ID: 2518
DOI: https://doi.org/10.15177/seefor.25-18
ORIGINAL SCIENTIFIC PAPER
Hierarchical Forest Site Classification According to the Perennial Plant Species in the Alacadağ Nature Reserve District
Kubilay Yatman1*, Serkan Gülsoy1
(1) Isparta University of Applied Sciences, Faculty of Forestry, Department of Forest Engineering, TR‑32260 Isparta, Turkey
Citation: Yatman K, Gülsoy S, 2025. Hierarchical Forest Site Classification According to the Perennial Plant Species in the Alacadağ Nature Reserve District.South-east Eur for 16(2): early view. https://doi.org/10.15177/seefor.25-18.
Received: 3 Feb 2025; Revised: 30 May 2025; Accepted: 28 Aug 2025; Published online: 16 Nov 2025
Cited by: Google Scholar
Abstract
In the study, vegetation was classified according to the distribution of perennial plant species in the Alacadağ Natural Reserve (NR) district which has high plant species richness. Classified vegetation groups were modeled by using environmental factors. Using these models, vegetation communities were mapped according to hierarchical classification. A grid network with a cell size of 100 · 100 m was created and 150 sample plots of 20 · 20 m in size were established. Plant species coverage was recorded in field studies according to the Braun-Blanquet scale. Cluster analysis was applied for the classification of vegetation communities based on binary hierarchical classification. Cluster groups were modeled with environmental variables using the classification tree analysis. The accuracy of the vegetation groups as a result of cluster analysis and modeling was tested by chi-square, kappa, and multiple permutation analyses at each distinction stage. Model groups with significant results were generalized and mapped on digital layers by using the prediction values. Indicator species for the vegetation groups were identified by indicator species analysis. As a result of this process, the study area was divided into 4 vegetation groups. The elevation and, correspondingly, climate were the most effective environmental variables in the differentiation of vegetation groups. In addition, ruggedness, hillshade, roughness, and heat index were the other important environmental variables for the vegetation groups in the district. As a result of this study, forest sites were classified for the conservation, sustainability, development, and future planning of the Alacadağ NR region.
Keywords: Alacadağ Natural Reserve; clustering; modeling; classification; vegetation
INTRODUCTION
The presence, growth, development, death, regeneration, and productivity of plants are the most fundamental elements of change and dynamism in forest areas (McDowell et al. 2020). This change and dynamism maintain its continuity under the influence of human activities and natural factors. Within the current dynamism, environmental variables enable the formation of vegetation elements such as trees, shrubs, grasses, etc. in forest areas. The fundamental method for characterizing the spatial distributions, diversities, structural characteristics, and productivity of plant species in forest succession stages is the identification of their relationships with environmental factors (Siddiqui et al. 2013). However, communities within forest ecosystems are highly complex and contain a lot of uncertain information that needs to be analysed. To reveal the correct information in this complex structure, communities first need to be classified and then ecologically related to environmental variables (Özkan and Negiz 2011). In other words, by associating living individuals or classified biotic communities with information pertaining to environmental variables in ecosystems, the complex interactions within the environment can be elucidated. For this purpose, it is imperative that detailed inventory studies be conducted primarily within ecosystems. During the inventory process, raw data pertaining to biotic and abiotic components are primarily collected from the field, followed by the processing of these data using various analytical techniques. This enables the acquisition of detailed and technical information about individuals and biotic communities.
It is technically possible to divide ecological research in forest sites into two subgroups: studies on autecology and on synecology (also known as community ecology) (Özkan et al. 2021). An important part of the studies completed in the field of community ecology in the literature consists of topics such as vegetation communities, biological diversity, and population behavior of living organisms. Vegetation classifications have an important place in these studies. In these studies, when vegetation-environment relationships are considered, the preferred dependent variables are perennial plant species. In this context, perennial plant species, which are non-motile organisms in their own sites, play a significant role in vegetation classification studies. Due to their longer life cycles, they exhibit greater spatial and temporal stability, making them reliable indicators of environmental gradients (Dengler et al. 2008, Kent 2011). Moreover, perennial species are less affected by short-term disturbances such as seasonal climatic fluctuations or episodic grazing (Kleyer et al. 2008), and they tend to persist even under moderate human or animal pressure. Additionally, compared to ephemeral annuals, perennials are often easier to identify during field surveys, as they are present for longer periods and retain diagnostic features more consistently (Chytrý et al. 2002, Özkan and Negiz 2011).
In vegetation classification studies, firstly, the recording of information on plant species (presence-absence or cover-abundance) is carried out according to the coordinates of the sampling sites through inventory studies in the forest area. To ensure the correct classification of vegetation, the execution of an accurate and successful inventory process is the most basic start (Grossman et al. 1998). In most studies on this subject, Braun-Blanquet (1932) method has been preferred (Chytrý and Otýpková 2003, Willner et al. 2009). Flora inventory data collected from different site conditions—once recorded manually—are now digitized and processed numerically, serving as a crucial input for vegetation classification. While analytical methods such as ordination and cluster analysis have long been used in phytosociology, the advent of computer-assisted multivariate techniques has significantly accelerated the analysis process and enabled the handling of large and complex datasets (De Cáceres and Wiser 2012). Thanks to these techniques, vegetation classification can be conducted hierarchically, while simultaneously identifying the distinctive plant species that are influential at each stage (Özkan et al. 2021). The most basic alternative methods that can be used to create vegetation groups objectively and analytically are association analysis (Williams and Lamberg 1959), cluster analysis (Pritchard and Anderson 1971) and two-way indicator species analysis (Hill 1979). These methods differ from each other by the use of alternative distance measurement formulas and by the detailed information they present in the result dendrograms.
After the classification of vegetation communities, the descriptive environmental variables for the resultant groups are identified. In this process, ordination methods, which commonly provide extensive and visual outputs, are frequently utilized (Fontaine et al. 2007, Özkan et al. 2009). It is possible to elaborate vegetation and environment relationships using various modeling techniques and to convert current model information into habitat or ecosystem classification maps using digital maps (Franklin 1995, Guisan and Zimmermann 2000, Elith and Leathwick 2009). These studies are generally referred to by different definitions, such as model-based ecosystem classification, ecological land classification, or forest site classification (Clare and Ray 2001, Bailey 2009). In the literature, it is also common to find studies where model-based vegetation-environment relationships are simulated and mapped according to climate change scenarios for future projections (Hickler et al. 2012, Hengl et al. 2018, Zhang et al. 2025, Zhao et al. 2025). In Türkiye, the number of studies that classify and map vegetation communities in forest areas is very limited (Özkan and Gülsoy 2010, Özkan and Mert 2011). It is known that a substantial portion of the existing studies do not employ modeling techniques. In other words, forest site classification studies that address model-based relationships for vegetation classification in Türkiye are almost non-existent (Özkan 2014). Although this is the case, the fact that the country's forest areas contain many different plants and living species, high biological diversity, complex land structure and geomorphological formations, as well as have a very high climatic variation, large forest areas and manage these areas sustainably make it obligatory to classify vegetation communities in our forest areas.
Forest site classification plays an important role in ecological planning and the sustainable use of forest resources. Considering the variation of plant species and vegetation communities in forest sites within ecosystems with a large and very heterogeneous geographical structure and variation in environmental factors, such as Türkiye, the results vary. Therefore, there is a growing need for forest site classification studies to be conducted at multiple spatial scales such as regional, local, and watershed levels to support ecologically relevant planning and adaptive forest management. At this stage, protected areas, particularly those with high ecological value or complex management needs, represent important forest units where forest site classification studies can provide significant contributions, both at national and international scales. To support the sustainability of the protected areas and to guide natural restoration efforts when needed, model-based forest site classification represents a valuable approach, particularly in data-driven contexts, even though it can be complemented by other ecological assessment and planning methods. This study, on this basis, was performed to classify, model and map vegetation communities according to perennial woody and herbaceous plant species in Alacadağ Nature Reserve in Finike district of Antalya province. The results from the study are aimed to provide important information regarding planning, protection and management of the area. The potential of this study to serve as a practical guide for places with protected area status on both national and international level is significant.
MATERIALS AND METHODS
Study Area
The study area, Alacadağ Nature Reserve (NR), which was designated as a protected site in 2017, is located between 36° 25' 00'' and 36° 18' 43'' north latitudes, and 30° 01' 30'' and 30° 05' 24'' east longitudes. Alacadağ NR is situated in the southwestern part of the Beydağları section, within the boundaries of the Finike and Demre districts in the western Mediterranean region of Türkiye (Figure 1). Although Alacadağ NR itself covers approximately 425 hectares, the study area was extended to a total of 4,012 hectares, including the surrounding forested zones. This broader scope was chosen to assess how ecologically distinct Alacadağ NR is from its adjacent vegetation, thereby enabling a more robust and meaningful forest site classification.
Figure 1. Study area: the Alacadağ NR district map .
Most of the mountains seen in a rugged structure in the area are in the form of an extension of the Western Taurus Mountains, and the highest point of the area is the Alacadağ with a height of 2,302 m (Öner and Vardar 2018). According to the Thornthwaite method (1948), the moisture effective index (Im) in the Alacadağ region has been determined to be 24.21. Based on this value, the prevailing climate type and precipitation effectiveness class in the region were characterized as “humid”.
The study area contains a large number of tree species, including rare forest tree species, and mostly reflects the characteristics of the Mediterranean floristic region in terms of vegetation diversity. Due to variables such as land surface forms, soil characteristics, and differences in elevation, it is possible to observe a wide variety of vegetation structures within the area. In the forests of the study area, the predominant species are Cedrus libani A. Rich var. libani, Juniperus excelsa M. Bieb subsp. excelsa, Juniperus foetidissima Willd, Pinus brutia Ten., and various deciduous tree species. Apart from forest vegetation, the area is dominated by bush vegetation in the form of maquis, steppe vegetation in high mountainous parts and rock vegetation with steep slopes where forest cover is not dominant. In addition, in some parts of the area, it is possible to come across aquatic vegetation structure and grassland, pasture or meadow vegetation. All these vegetation differences mentioned in the area enrich the region in terms of habitat diversity, ecological diversity and species diversity.
Land Survey and Data Preparation
Digital elevation model (DEM) was created according to the area boundaries by using the contour curves of the 1/25,000 scale topographic maps (P24-A1-95 and P24-A4-95). In the inventory process, perennial woody and herbaceous plant species were recorded in 150 sample plots (20 · 20 m), distributed according to a stratified sampling design. Strata were defined based on key environmental gradients, including elevation (low, mid, high zones), slope classes (gentle, moderate, steep), and dominant vegetation types (forest, bush/maquis, steppe). The plots were proportionally allocated within each stratum and positioned to maximize coverage while avoiding spatial clustering. The coordinates of all sample plots were recorded as in the UTM WGS84 coordinate system using the Global Position System (GPS). In each of these areas distributed between 374 and 1,866 m, the abundance and cover of perennial plant species were visually assessed and recorded using the semi-quantitative Braun-Blanquet (1932) cover-abundance scale. Then, this data set was converted into binary data (presence-absence) and prepared for statistical analysis.
Following the field studies, the previously created DEM was used in ArcGIS, and elevation, slope, ruggedness, roughness and hillshade index, and topographic position index maps of Alacadağ NR were created, respectively (Jenness 2006). The radiation index and heat index maps were created by calculating with the relevant formulas (Moisen and Frescino 2002, Aertsen et al. 2010, Wei et al. 2010, Brown and Ahl 2011). For climate maps 19 different bioclimatic variables (750 m resolution) by Hijmans et al. (2005) were downloaded in ASCII format, cut at the scale of the study area and made ready for use. The coordinates of 150 sample plots were transferred to the digital bases and the data were obtained for all descriptive environmental variables. These variables were given codes and prepared for the statistical analysis (Table 1).
Table 1. The environmental variables included in the study and their abbreviations.
Statistical Evaluation
Within the scope of the research, the vegetation classification approach was applied under the principle of binary analytic hierarchical distinction (Özkan et al. 2013). At this stage, the vegetation data matrix was evaluated by cluster analysis (McCune and Grace 2002). In the cluster analysis, the Jaccard-Ward method was applied in the selection of the combination of distance measurement and group distinction technique. The groups identified through cluster analysis were recorded as class variables and modeled with environmental variables using the classification tree method (De'ath and Fabricius 2000). Receiver Operating Characteristic (ROC) analysis was applied to test the performance of the models (Hanley and McNeil 1982). Bioclimatic variables were examined for multicollinearity using Pearson correlation analysis (Özdamar 2013). Variable pairs with high correlation coefficients (r ≥ 0.80) were considered redundant, and in such cases, the variable with lower ecological relevance or interpretability was excluded from further analysis. In order to determine the degree of agreement between vegetation groups obtained hierarchically through cluster analysis and model groups, chi-square tests (Cole 1949), Cohen's kappa statistics (Cohen 1960, Viera and Garrett 2005) and Multi-Response Permutation Procedure (MRPP) test (Zimmerman et al. 1985) were applied at each distinction stage, respectively (Poole 1974, Ozkan 2002). Indicator plant species of vegetation groups with statistically significant results were identified through indicator tests (Dufrene and Legendre 1997, Negiz et al. 2015). Following these analyses, ArcGIS software was utilized to map forest site classes in a hierarchical order. Statistical analyses within the scope of the study were performed in open-source R-Studio using AUC (Ballings and Van den Poel 2013), cluster (Maechler 2013), factoextra (Kassambara and Mundt 2020), dendextend (Galili 2015), caret (Kuhn et al. 2020), chisq. posthoc.test (Ebbert 2019), corrplot (Wei and Simko 2017), ROCR (Sing et al. 2005), and tree (Ripley and Ripley 2016) packages. Additionally, PCOrd software was used to conduct the Multi-Response Permutation Procedure (MRPP) to test the statistical significance of group differentiation, while DTREG software was used to perform indicator species analysis to identify diagnostic species associated with specific forest site types.
RESULTS
As a result of the inventory studies carried out in 150 sample plots, 60 different tree and shrub species and 8 perennial herbaceous species (with distinctive forest site characteristics) were included in the vegetation data matrix in the study. Considering the frequency of these plant species in the sample plots, 37 of the perennial woody species and 4 of the perennial herbaceous species had frequency values greater than 3%. From this point of view, the vegetation data matrix created with a total of 41 plant species was subjected to analytical evaluation for forest site classification in the study (Table 2).
Table 2. Plants to be statistically evaluated and their abbreviations.
In the study, a Pearson correlation analysis was initially conducted to identify descriptive variables with high correlation (r > 0.85) that could lead to multicollinearity issues in the models. As a result of the analysis, all climate variables except for bio2 and bio12 were found to be highly correlated with elevation (Elvtn). Therefore, a total of 10 variables were chosen for the modeling processes: 8 topographical (Elvtn, HeaInd, Hillsh, RadInd, RoughI, RggedI, Slope, TPI) and 2 bioclimatic (bio2, bio12).
According to the cluster analysis applied based on the hierarchical vegetation distinction, the vegetation classification of Alacadağ NR was completed in 3 stages. The vegetation groups obtained by cluster analysis were modeled with the classification tree technique with environmental variables as dependent variables at each distinction stage. The ROC (receiver operating characteristic) values of the training and test data sets at each distinction stage and the descriptive variables in the models are given in Table 3.
Cohen's kappa statistics, chi-square values, and their significance levels (p < 0.05) calculated based on the number of sample plots that have deviated and shifted between the vegetation groups obtained from the cluster analysis and the model groups formed as a result of the modeling analysis are presented in Table 4.
As a result of the cluster analysis and classification tree models, the model groups that were significantly (p < 0.05) separated from each other were classified into four different forest site classes, coded as MG1, MG2-1, MG3-1, and MG3-2 (Figure 2).
Then, a Multi-Response Permutation Procedure (MRPP) test was used to determine intergroup distance (T) and intragroup homogeneity (A) values in these model groups at each distinction stage (Table 5).
According to the results of the MRPP test, the distance of the model groups to each other was the highest at the first distinction stage, while the intra-group homogeneity value was the highest at the second distinction stage. In the next part, the distinctive indicator plant species of these forest site classes at the local scale were determined. Consequently, the hierarchical site classification of forest areas in Alacadağ NR was completed, and a final map containing indicator species for each forest site class was obtained (Figure 3).
In Alacadağ NR, four different forest site classes were obtained at three distinction stages. At the end of the first distinction stage, the variables of elevation, ruggedness index and hillshade index were decisive for the formation of forest site classes. According to the model rules, Alacadağ NR was classified into a separate forest site named Model Group 1 (MG1) based on the following criteria: a) areas with an elevation above 1,538.5 meters and a ruggedness index greater than 0.0428, and b) areas with an elevation between 1,118.5 meters and 1,538.5 meters, with a ruggedness index greater than 0.0428, and a hillshade index greater than 109.5. Correspondingly, Model Group 2 (MG2) was classified based on the following criteria: a) areas with an elevation less than or equal to 1,188.5 meters, b) areas with an elevation greater than 1,188.5 meters and a ruggedness index less than or equal to 0.0428, and c) areas with an elevation between 1,188.5 meters and 1,538.5 meters, with a ruggedness index greater than 0.0428, and a hillshade index less than or equal to 109.5.
The MG1 represents Cedrus libani forests of the district, which are generally between 1,100 and 1,500 m, in shady areas with a lot of ruggedness, and more rugged areas above 1,500 m. In addition, important distinguishing plants in this habitat class were Acer hyrcanum subsp. sphaerocaryum, Juniperus excelsa, Juniperus foetidissima, and Ostrya carpinifolia Scop. MG2, on the contrary, represents general forest sites in lower hilly and less shaded areas between 1,100 and 1,500 m in elevation. These areas generally correspond to the description of the Pinus brutia forest sites in Alacadağ NR, extending from the lowest elevation to the highest elevation limit. In MG2, the most dominant distinguishing species are Quercus coccifera L., Pistacia terebinthus L., Arbutus andrachne L., Cotinus coggygria Scop., Cistus creticus L., and Phillyrea latifolia L., which are typical indicators of the Mediterranean phytogeography. At the end of the first classification stage, MG1 in Alacadağ Nature Reserve was not further subdivided into another class, while MG2 was further divided into two distinct forest site classes coded as MG2-1 and MG2-2.
MG2, as described above, generally represents forest areas below 1,500 m. In MG2, ruggedness index and elevation variables were decisive in the distinction of forest site classes. According to the model rules at this stage, in Alacadağ NR areas with an elevation greater than 1,530.5 meters and a ruggedness index greater than 0.0365 were classified as a separate forest site class, MG2-1. The indicator plant species of the MG2-1 forest site were found to be quite similar to those of the MG1 forest site, which is classified at a higher level. The most dominant distinguishing species that set MG2-1 apart from MG1 are Origanum onites L., Verbascum sinuatum L., Paliurus spina-christi P. Mill., Prunus divaricata, and Celtis australis. The model rules defining the MG2-2 forest site class within the MG2 category are as follows: a) areas with an elevation less than or equal to 1,530.5 meters, and b) areas with an elevation greater than 1,530.5 meters where the ruggedness index is less than or equal to 0.0365. The MG2-2 forest site is represented by the association of tree and shrub species including Pinus brutia, Quercus coccifera, Pistacia terebinthus, Cistus creticus, Cotinus coggygria, Arbutus andrachne, and Genista acanthoclada DC. In this forest site class, Genista acanthoclada is included as a different distinguishing species, while Phillyrea latifolia, which is present in the MG2 forest site, is not found here. At the end of the second classification stage, MG2-1 in Alacadağ NR was not further subdivided into another class, while MG2-2 was further divided into two distinct forest site classes coded as MG3-1 and MG3-2.
In the third classification stage, the variables that influenced the determination of forest site classes in Alacadağ Nature Reserve were, in order, ruggedness index, heat index, and elevation. According to the new model rules obtained within MG2-2, the following were classified as a separate forest site class coded as MG3-1: a) areas where the ruggedness index is greater than 11.698, and b) areas where the ruggedness index is less than or equal to 11.698, the heat index is greater than 0.842, and the elevation is greater than 861.5 meters. The dominant distinguishing species of this forest site are Cedrus libani, Cotinus coggygria, Cercis siliquastrum L., Quercus coccifera, Fraxinus ornus, and Laurus nobilis L. Based on this information, it is understood that the rugged terrain structure shaping from the lower elevations to the upper elevation steps of Alacadağ NR is the most defining environmental descriptor for this forest site. According to the other model rules at this stage, a) areas where the ruggedness index is less than or equal to 11.698 and the heat index is less than or equal to 0.842, and b) areas where the ruggedness index is less than or equal to 11.698, the heat index is greater than 0.842, and the elevation is less than or equal to 861.5 meters, were classified as the forest site class coded as MG3-2. For MG3-2, the most prominent distinguishing species are Origanum onites, Smilax aspera L., Genista acanthoclada, Phillyrea latifolia, Verbascum sinuatum, Thymelaea tartonraira (L.) All., and Olea europaea L. In the shaping of this forest site, the parts where the heat index is high at the lowest elevation step of the region were decisive.
DISCUSSION
Increasing demands on forest ecosystem services such as timber supply, carbon storage, biodiversity, and recreation driven by continued population growth and land-use change, are placing escalating pressure on forest areas globally (Dublin et al. 2004, Sloan and Sayer 2015, Curtis et al. 2018, FAO 2022). On the other hand, forests are being destroyed as a consequence of global threats such as climate crisis and environmental pollution, whose effects we can already experience today. Given the widely accepted prediction that the Mediterranean basin is one of the most vulnerable regions to the climate crisis characterized by rising temperatures, prolonged droughts, and increasing wildfire risks, monitoring the status of Türkiye’s Forest ecosystems, particularly of the protected areas, has become increasingly important (Cramer et al. 2018, Lionello and Scarascia 2018, Lee et al. 2023). The results of this study demonstrate that vegetation classification based on perennial woody and herbaceous species, combined with environmental modeling, is an effective approach for defining forest site classes within protected areas such as Alacadağ Nature Reserve. By identifying ecologically distinct vegetation communities and mapping their spatial distribution, we have provided a practical foundation for monitoring ecological changes and supporting forest planning efforts under the principles of sustainability (Guisan and Theurillat 2000, Fontaine et al. 2007, Özkan 2009).
The identification of forest site classes that host high species diversity, endemic or relict taxa, and ecologically valuable resources such as medicinal and aromatic plants, plays a crucial role in determining priority areas for conservation and sustainable management. These findings are particularly relevant given the increasing threats to natural forest ecosystems due to climate change, pollution, forest pests, invasive species, fires, and grazing pressure (Başkent et al. 2003, Özkan and Gülsoy 2010, Özkan and Mert 2011). The generated maps can serve as baseline tools for forest managers to track ecosystem dynamics, assess site-specific pressures, and support regionally adaptive decisions.
In a global context where the loss of natural forest areas is accelerating, particularly in biodiversity hotspots like the Mediterranean region, the strategic conservation of intact forest ecosystems has become increasingly important (Bruijnzeel 2004, Ticktin 2004). Our study reinforces the notion that protected forest sites serve as crucial gene pools and refugia for endemic and rare species, while simultaneously providing essential ecosystem services such as hydrological regulation, soil protection, bioenergy potential, and carbon sequestration.
Therefore, the integrative approach employed in this study—linking vegetation to environmental gradients and generating model-based site classification maps—contributes valuable insights for modern forestry practices. These outputs support both conservation and utilization objectives, providing actionable data for long-term ecological monitoring and sustainable resource use in Alacadağ NR and similar protected areas.
As a result of all the applied classification and modeling processes, the Alacadağ NR district was divided into four different forest site classes and mapped accordingly. Similar studies have been carried out in Türkiye and abroad for the purpose of the planning of forest ecosystems, the improvement of destroyed forest communities and the conservation of natural resources (Brzeziecki et al. 1993, Zimmermann and Felix 1999, Miller and Franklin 2002, Fontaine et al. 2007, Liu et al. 2009, Özkan and Negiz 2011). Since plant species characterize the features of their habitats, the obtained model-based vegetation distribution maps are an important ecological information pool. In the case of forest sites they represent and serve as a guide for planners. In other words, the classification of vegetation communities and forest sites is important for the planning of forest ecosystems accurately in the future. On the other hand, with the information from such studies, the future status of species and vegetation communities during a possible climate change can be predicted (Thomas et al. 2004). Therefore, if future climate scenarios are included in such models, it is possible to determine the effects of climate change on vegetation communities or the distribution of species.
In light of the provided information, this study modeled the relationships between the distribution of perennial woody and herbaceous species and environmental variables in a hierarchical manner. Considering all the results, the environmental variables that were effective in the distinction of the forest sites were elevation, hillshade and ruggedness index, heat index and roughness index. In general, the elevation variable was the most effective in the distinction of the groups. In general, elevation was the most influential environmental variable in the classification of forest site types in Alacadağ NR. It can be said that all other climatic factors, primarily temperature and precipitation, which vary with elevation, are also decisive here. This indicates that potential future climate change will have a significant impact on such protected areas. Therefore, to manage the process most accurately in these areas, there is a need for up-to-date and modern conservation action plans. In this context, model-based forest site classification and maps are among the most important guiding resources.
The final forest site classification map obtained in this study has significant advantages compared to maps produced by the classical approach. The first of these advantages is that the proximity of the different forest site units separated on the map can be determined much more clearly. In other words, it is much easier to perceive how close or distant these units are to each other in terms of their ecological characteristics. For example, in this study, the similarity between the MG1 forest site from the first classification stage and the MG2-1 forest site from the second classification stage was determined using this approach. The hillshade index has been particularly decisive in the distinction of these sites. The MG2-2 vegetation group, which has been distinguished within the MG2 forest site, is characterized by typical Red Pine forests (Pinus brutia) and warm Mediterranean and maquis elements. The first distinction within this vegetation group, MG3-1, represents areas above approximately 850 meters and extending to the upper limit of Pinus brutia in sunny aspects. In the lower elevations of this forest site, species such as Cotinus coggygria, Laurus nobilis, Cercis siliquastrum, and Quercus infectoria are identified, while distinguishing species such as Cedrus libani and Fraxinus ornus are found at higher elevations. The MG3-2 vegetation group has revealed a forest site characterized by warm Mediterranean, maquis, or garrigue communities dominated by distinguishing species such as Origanum onites, Smilax aspera, Genista acanthoclada, Phillyrea latifolia, Verbascum sinuatum, and Olea europaea. These species are generally distributed below approximately 850 meters in Alacadağ NR, in areas with less rugged terrain compared to higher elevations.
In conclusion, this study’s results are of great importance in terms of determining the regional limits of seed transfer between similar growing environments or determining strategies to increase the diversity in the area through distinctive species. In this way, appropriate strategies can be determined for more effective conservation of the area under changing environmental conditions. The tangible outcomes obtained from the study can be utilized in the process of taking measures to protect species against factors that may have adverse effects on the ecosystem, such as the climate crisis, fire, insect damage, invasive species, and environmental pollution. All the findings in this study were determined by using analytical methods. With the information obtained from these methods, the boundaries of the forest site types identified in the area can be monitored in the future. In this way, it will be possible to observe the effects of adverse events such as the climate crisis on the natural site in the future. Additionally, this study provides indirect information for important ecosystem functions such as wildlife, erosion prevention, clean water, the conservation of endemic, rare, and relic plant species, and the use of medicinal aromatic plants in the district. In summary, when all the results revealed in the study are taken into consideration, it is concluded that the study can facilitate both scientific and practical applications.
CONCLUSIONS
This study highlights the crucial role of integrating vegetation-based classification with environmental modeling for the sustainable management of forest ecosystems, particularly within sensitive and protected Mediterranean landscapes such as Alacadağ Nature Reserve. The results demonstrate that elevation and topographical heterogeneity are key determinants of forest site differentiation, shaping the distribution of both woody and herbaceous perennial species. By linking floristic composition with environmental gradients, this approach offers a reproducible and scalable framework for ecological monitoring and adaptive forest planning under changing climatic conditions. Moreover, the study underscores the potential of model-based forest site maps as strategic tools to guide restoration, seed transfer, and conservation practices. Such integrative analyses contribute not only to the scientific understanding of forest ecology but also to the formulation of modern, data-driven conservation policies aimed at preserving biodiversity, mitigating climate impacts, and ensuring the long-term resilience of Mediterranean forest ecosystems.
Author Contributions
KY and SG conceived and designed the research, KY carried out the field measurements and data collection, KY performed the statistical analyses and modelling, SG supervised the research, contributed to data interpretation, and reviewed the manuscript, KY prepared the manuscript draft, and SG provided critical revisions, both authors read and approved the final version of the manuscript.
Funding
This study was supported by Project No. 2019-YL1-0038 funded by Isparta University of Applied Sciences Scientific Research Projects Management Unit. In addition, the knowledge obtained from the courses numbered 1129B372000372, 1129B371801389 and 1129b371801398 supported by TÜBİTAK 2237-A was used in the study.
Conflicts of Interest
The authors declare no conflict of interest.
REFERENCES
Aertsen W, Kint V, Van Orshoven J, Özkan K, Muys B, 2010. Comparison and ranking of different modelling techniques for prediction of site index in Mediterranean mountain forests. Ecol Model 221(8): 1119-1130. https://doi.org/10.1016/j.ecolmodel.2010.01.007.
Ballings M, Van den Poel D, 2013. AUC: Threshold independent performance measures for probabilistic classifiers. R package version 0.3.0. Available online: https://www.r-project.org/ /package= AUC (08 November 2024).
Başkent EZ, Barlı Ö, Ayaz H, Bilgili E, Turna İ, İpek A, Altun L, 2003. A different approach to the restructuring of Turkish forestry (I). Orman ve Av, 6, Türkiye, Ankara.
Bailey RG, 2009. Ecosystem geography: From ecoregions to sites. Second edition. Springer, New York, USA, 251 p. https://doi.org/10.1007/978-0-387-89516-1.
Braun-Blanquet J, 1932. Plant Sociology; The study of plant communities. (Eds Fuller GD and Conard HS). McGraw-Hill book company, New York and London, 439 p.
Brown Jr S, R. & Ahl, R, S, 2011. The region 1 Existing Vegetation Mapping Program (Vmap) Beaverhead-Deerlodge Methodology. Region One Vegetation Classification, Mapping, Inventory and Analysis Report No:11-02: 1-18.
Bruijnzeel LA, 2004. Hydrological functions of tropical forests: not seeing the soil for the trees. Agr Ecosyst Environ 104(1): 185-228. https://doi.org/10.1016/j.agee.2004.01.015.
Brzeziecki B, Kienast F, Wildi O, 1993. A simulated map of the potential natural forest vegetation of Switzerland. J Veg Sci 4(4): 499-508. https://doi.org/10.2307/3236077.
Chytrý M, Otýpková Z, 2003. Plot sizes used for phytosociological sampling of European vegetation. J Veg Sci 14(4): 563-570. https://doi.org/10.1111/j.1654-1103.2003.tb02183.x.
Chytrý M, Tichý L, Holt J, Botta‐Dukát Z, 2002. Determination of diagnostic species with statistical fidelity measures. J Veg Sci 13(1): 79-90. https://doi.org/10.1111/j.1654-1103.2002.tb02025.x.
Clare J, Ray D, 2001. A spatial model of Ecological Site Classification for forest management in Britain. In Proceedings of the 4th AGILE conference on Geographic Information Science, pp. 19-21.
Cohen J, 1960. A coefficient of agreement for nominal scales. Educ Psychol Meas 20(1): 37-46. https://doi.org/10.1177/001316446002000104.
Cole LC, 1949. The measurement of interspesific association. Ecology 30(4): 411-424. https://doi.org/10.2307/1932444.
Cramer W, Guiot J, Fader M, Garrabou J, Gattuso JP, Iglesias A, Lange MA, Lionello P, Llasat MC, Paz S, Peñuelas J, et al., 2018. Climate change and interconnected risks to sustainable development in the Mediterranean. Nat Clim Change 8(11): 972-980. https://doi.org/10.1038/s41558-018-0299-2.
Curtis PG, Slay CM, Harris NL, Tyukavina A, Hansen MC, 2018. Classifying drivers of global forest loss. Science 361(6407): 1108-1111. https://doi.org/10.1126/science.aau3445.
De'ath G, Fabricius KE, 2000. Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 81(11): 3178-3192. https://doi.org/10.1890/0012-9658.
De Cáceres M, Wiser SK, 2012. Towards consistency in vegetation classification. J Veg Sci 23(2): 387‑393. https://doi.org/10.1111/j.1654-1103.2011.01354.x.
Dengler J, Chytry M, Ewald J, 2008. Phytosociology. In: Jørgensen SE, Fath BD (eds) Encyclopedia of ecology. Academic Press, Elsevier, Oxford, UK, pp. 2767–2779.
Dublin HT, Volonte C, Brann J, 2004. GEF Biodiversity Program Study, 2004. Global Environment Facility, Office of Monitoring & Evaluation, 131 p.
Dufrene M, Legendre P, 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecol Monogr 67: 345-366. https://doi.org/10.1890/0012-9615.
Ebbert D, 2019. A post hoc analysis for Pearson’s chi-squared test for count data. R package version 0.1.2.
Elith J, Leathwick JR, 2009. Species distribution models: ecological explanation and prediction across space and time. Annu Rev Ecol Evolut Syst 40(1): 677-697. https://doi.org/10.1146/annurev.ecolsys.110308.120159.
FAO T, 2022. The state of the world’s forests 2022. Forest pathways for green recovery and building inclusive, resilient and sustainable economies. FAO, Rome, Italy, 166 p.
Fontaine M, Aerts R, Özkan K, Mert A, Gülsoy S, Süel H, Waelkens M, Muys B, 2007. Elevation and exposition rather than soil types determine communities and site suitability in Mediterranean mountain forests of southern Anatolia, Türkiye. Forest Ecol Manag 247: 18-25. https://doi.org/10.1016/j.foreco.2007.04.021.
Franklin J, 1995. Predictive vegetation mapping: geographic modelling of biospatial patterns in relation to environmental gradients. Prog phys geog 19(4): 474-499. https://doi.org/10.1177/030913339501900403.
Galili T, 2015. dendextend: an R package for visualizing, adjusting and comparing trees of hierarchical clustering. Bioinformatics 31(22): 3718-3720. https://doi.org/10.1093/bioinformatics/btv428.
Guisan A, Theurillat J, 2000. Assessing alpine plant vulnerability to climate change: a modelling perspective. Integrated Assessment 1: 307-320. https://doi.org/10.1023/A:1018912114948.
Guisan A, Zimmermann NE, 2000. Predictive habitat distribution models in ecology. Ecol Model 135(2‑3): 147-186. https://doi.org/10.1016/S0304-3800(00)00354-9.
Hanley JA, McNeil BJ, 1982. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143(1): 29-36. https://doi.org/10.1148/radiology.143.1.7063747.
Hengl T, Walsh MG, Sanderman J, Wheeler I, Harrison SP, Prentice IC, 2018. Global mapping of potential natural vegetation: an assessment of machine learning algorithms for estimating land potential. PeerJ 6: e5457. https://doi.org/10.7717/peerj.5457.
Hickler T, Vohland K, Feehan J, Miller PA, Smith B, Costa L, Sykes MT, 2012. Projecting the future distribution of European potential natural vegetation zones with a generalized, tree species‐based dynamic vegetation model. Global Ecol Biogeogr 21(1): 50-63. https://doi.org/10.1111/j.1466-8238.2010.00613.x.
Hill MO, 1979. TWINSPAN -a Forttan program for arranging multivariate data in an ordered two way table by classification of the individuals and attributes. Cornell University, Ithaca, New York, USA, 90 p.
Jenness J, 2006. Topographic position index (tpi_jen. avx) extension for ArcView 3. x version 1.2. Jenness Enterprises, Flagstaff, AZ. Available online: http://www.jennessent.com/arcview/tpi.htm (17 June 2024).
Kassambara A, Mundt F, 2020. factroextra: Extract and Visualize the Results of Multivariate Data Analyses. R package version 1.0.7. Available online: https://CRAN.R-project.org/package=factoextra (24 January 2025).
Kent M, 2011. Vegetation description and data analysis: a practical approach. John Wiley & Sons, UK, 448 p.
Kleyer M, Bekker RM, Knevel IC, Bakker JP, Thompson K, Sonnenschein M, Poschlod P, Van Groenendael JM, Klimeš L, Klimešová J, Klotz SRGM, et al., 2008. The LEDA Traitbase: a database of life‐history traits of the Northwest European flora. J ecol 96(6): 1266-1274. https://doi.org/10.1111/j.1365-2745.2008.01430.x.
Kuhn M, Wing J, Weston S, Williams A, Keefer C, Engelhardt A, Cooper T, Mayer Z, Kenkel B, Team RC, 2020. Package ‘caret’. The R Journal, 223(7), 48.
Lee H, Calvin K, Dasgupta D, Krinmer G, Mukherji A, Thorne P, Trisos C, Romero J, Aldunce P, Barret K, Blanco G, 2023. Synthesis report of the IPCC Sixth Assessment Report (AR6), Longer report. IPCC.
Lionello P, Scarascia L, 2018. The relation between climate change in the Mediterranean region and global warming. Reg Environ Change 18: 1481-1493. https://doi.org/10.1007/s10113-018-1290-1.
Liu H, Wang L, Yang J, Nakagoshi N, Liang C, Wang W, Lv Y, 2009. Predictive modelling of the potential natural vegetation pattern in northeast China. Ecol Res 24: 1313-1321. https://doi.org/10.1007/s11284-009-0616-3.
Maechler M, 2013. Cluster analysis extended Rousseeuw et al. R CRAN.
McCune B, Grace JB, 2002. Analysis of ecological communities. MJM Software Design, Gleneden Beach, Oregon, USA, 300 p.
McDowell NG, Allen CD, Anderson-Teixeira K, Aukema BH, Bond-Lamberty B, Chini L, (…), Xu C, 2020. Pervasive shifts in forest dynamics in a changing world. Science 368(6494): eaaz9463. https://doi.org/10.1126/science.aaz9463.
Miller J, Franklin J, 2002. Modelling the distribution of four vegetation alliances using generalized linear models and classification trees with spatial dependence. Ecol Model 157(2): 227-247. https://doi.org/10.1016/S0304-3800(02)00196-5.
Moisen GG, Frescino TS, 2002. Comparing five modelling techniques for predicting forest characteristics. Ecol Model 157(2): 209-225. https://doi.org/10.1016/S0304-3800(02)00197-7.
Negiz M, Gülsoy S, Özkan K, 2015. A land inventory approach that can be used to identify species diversity components in forest ecosystems. Süleyman Demirel University Journal of Graduate School of Natural and Applied Sciences 19(2): 198-204.
Öner E, Vardar S, 2018. Holocene geomorphology of the Finike plain and geoarchaeology of Limyra. Eurasian Journal of Social and Economic Research 5(5): 1-23.
Özdamar K, 2013. Analysis of statistical data with package programs (Volume 1). Ankara: Nisan Kitapevi, pp. 27-36.
Özkan K, Gülsoy S, 2010. Ecological land classification and mapping based on vegetation-environment hierarchical analysis - a case study of Buldan forest district (Turkey). Pol J Ecol 58(1): 55-67.
Özkan K, Mert A, 2011. Ecological land classification and mapping of Yazili Canyon Nature Park in the Mediterranean region, Turkey. J Environ Eng Landsc 19(4): 296-303. https://doi.org/10.3846/16486897.2011.638214.
Özkan K, Negiz MG, 2011. Classification and mapping of woody vegetation in Isparta Yukarıgökdere Region by hierarchical methods. SDU Journal of Forestry Faculty 1: 27-33.
Ozkan K, 2002. Measurement of interspecies coexistence by interspecific correlation analysis. Süleyman Demirel University Journal of Forestry Faculty A (2): 71-78.
Özkan K, 2009. Environmental factors as influencing vegetation communities in Acipayam district of Turkey. J Environ Biol 30(5): 741-746.
Özkan K, 2014. Hierarchical modelling based ecological land classification in a forest district of Mediterranean region, Turkey. Environ Eng Manag 4: 979-990.
Özkan K, Makineci E, Gülsoy S, Özdemir S, 2021. Forestry and forestry in Türkiye on the axis of ecology and economy. Soncag Academy, Ankara, Turkey pp. 359-406.
Özkan K, Mert A, Aertsen W, Muys B, 2013. Hierarchical land classification and mapping of Ağlasun Forest Ecosystems in the Mediterranean Region, Turkey. J Environ Biol 34(3): 623-633.
Özkan K, Şenol H, Gulsoy S, Mert A, Suel H, Eser Y, 2009. Vegetation‐environment relationships in Mediterranean mountain forests on limeless bedrocks of southern Anatolia, Turkey. J Environ Eng Landsc 17(3): 154-163. https://doi.org/10.3846/1648-6897.2009.17.154-163.
Poole RW, 1974. An introduction to quantitative ecology. McGraw-Hill, Inc., New York, USA, 532 p.
Pritchard NM, Anderson AJB, 1971. Observation on the use of cluster analysis in botany with an ecological example. J Ecol 59(3): 727-747. https://doi.org/10.2307/2258136.
Ripley B, Ripley MB, 2016. Package ‘tree’. Classification and Regression Trees. Version 1.0.0.
Siddiqui MF, Shaukat SS, Ahmed M, Khan N, Khan IA, 2013. Vegetation-environment relationship of conifer dominating forests of moist temperate belt of Himalayan and Hindukush regions of Pakistan. Pak J Bot 45(2): 577-592.
Sing T, Sander O, Beerenwinkel N, Lengauer T, 2005. ROCR: visualizing classifier performance in R. Bioinformatics 21(20): 3940-3941. https://doi.org/10.1093/bioinformatics/bti623.
Sloan S, Sayer JA, 2015. Forest Resources Assessment of 2015 shows positive global trends but forest loss and degradation persist in poor tropical countries. Forest Ecol Manag 352: 134-145. https://doi.org/10.1016/j.foreco.2015.06.01.
Thomas CD, Cameron A, Green RE, Bakkenes M, Beaumont LJ, Collingham YC, Erasmus BF, De Siqueira MF, Grainger A, Hannah L, et al., 2004. Extinction risk from climate change. Nature 427(6970): 145‑148. https://doi.org/10.1038/nature02121.
Thornthwaite CW, 1948. An approach toward a rational classification of climate. Geogr rev 38(1): 55‑94. https://doi.org/10.2307/210739.
Ticktin T, 2004. The ecological implications of harvesting non-timber forest products. J Appl Ecol 41(1): 11-21. https://doi.org/10.1111/j.1365-2664.2004.00859.x.
Viera AJ, Garrett JM, 2005. Understanding interobserver agreement: the kappa Statistic. Fam Med 37(5): 360-363.
Wei T, Simko V, 2017. R package “corrplot”: Visualization of a Correlation Matrix (Version 0.84). Available online: https://github.com/taiyun/corrplot (12 November 2024).
Wei XZ, Jiang MX, Huang HD, Yang JY, Yu J, 2010. Relationships between environment and mountain riparian plant communities associated with two rare tertiary-relict tree species, Euptelea pleiospermum (Eupteleaceae) and Cercidiphyllum japonicum (Cercidiphyllaceae). Flora - Morphology, Distribution, Functional Ecology of Plants 205: 841-852. https://doi.org/10.1016/j.flora.2010.04.003.
Willner W, Tichý L, Chytrý M, 2009. Effects of different fidelity measures and contexts on the determination of diagnostic species. J Veg Sci 20(1): 130-137. https://doi.org/10.1111/j.1654-1103.2009.05390.x.
Williams WT, Lambert JM, 1959. Multivariate methods in plant ecology – I. Association analysis in plant communities. J Ecol 47(1): 83-101. https://doi.org/10.2307/2257249.
Yalçın E, Bilgin A, Kutbay HG, Kılınç M, 2004. Relationships between community structure and soil properties of a swamp forest from northern Turkey. Pol J Ecol 14(2): 114-127.
Zhang H, Hu Z, Chen X, Li J, Zhang Q, Zheng X, 2025. Global Greening major contributed by climate change with more than two times rate against the history period during the 21th Century. Glob Change Biol 31(3): e70126. https://doi.org/10.1111/gcb.70126360.
Zhao S, Chen H, Zhang X, Xiao P, Bai L, 2025. VegeDiff: Latent diffusion model for geospatial vegetation forecasting. IEEE Transactions on Geoscience and Remote Sensing.
Zimmerman GM, Goetz H, Mielke Jr PW, 1985. Use of an improved statistical method for group comparisons to study effects of prairie fire. Ecology 66: 606-611. https://doi.org/10.2307/1940409.
Zimmermann NE, Felix K, 1999. Predictive mapping of alpine grasslands in Switzerland: Species versus community approach. J Veg Sci 10: 469-482. https://doi.org/10.2307/3237182.
© 2025 by the Croatian Forest Research Institute. This is an Open Access paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0).