Abstract
Global climate change poses significant threats to ecosystems worldwide, particularly impacting long-lived forest tree species such as Pinus nigra. This study assessed the potential shifts in distribution areas for Pinus nigra, an important tree species, one highly vulnerable to global climate change, given its prevalence in continental climates, in Türkiye under different climate scenarios (SSPs 585 and 245). In this study, suitable distribution regions of Pinus nigra were evaluated based on SSPs 585 and SSPs 245 using nine different models. Results indicated potential losses in Pinus nigra distribution areas ranging from 15.0% to 43.5% (SSPs 245) and 19.7% to 48.9% (SSPs 585) by 2100. However, in 2100, new suitable distribution areas are expected to be formed at rates ranging from 13.8% to 32.1% and 15.1% to 34.4% according to the above scenarios. Because most of the newly formed suitable distribution regions are quite far from the areas where the species currently spreads, it seems necessary to provide the migration mechanism needed by the species by humans to prevent population losses in this process.
Download PDF
Full Article
Possible Changes of Pinus nigra Distribution Regions in Türkiye with the Impacts of Global Climate Change
Uğur Cantürk,a,* İsmail Koç,b Halil Barış Özel,c and Hakan Şevik d
Global climate change poses significant threats to ecosystems worldwide, particularly impacting long-lived forest tree species such as Pinus nigra. This study assessed the potential shifts in distribution areas for Pinus nigra, an important tree species, one highly vulnerable to global climate change, given its prevalence in continental climates, in Türkiye under different climate scenarios (SSPs 585 and 245). In this study, suitable distribution regions of Pinus nigra were evaluated based on SSPs 585 and SSPs 245 using nine different models. Results indicated potential losses in Pinus nigra distribution areas ranging from 15.0% to 43.5% (SSPs 245) and 19.7% to 48.9% (SSPs 585) by 2100. However, in 2100, new suitable distribution areas are expected to be formed at rates ranging from 13.8% to 32.1% and 15.1% to 34.4% according to the above scenarios. Because most of the newly formed suitable distribution regions are quite far from the areas where the species currently spreads, it seems necessary to provide the migration mechanism needed by the species by humans to prevent population losses in this process.
DOI: 10.15376/biores.19.3.6190-6214
Keywords: Black pine; Continental climate; Modelling; ShinyBIOMOD; SSPs 245; SSPs 585
Contact information: a: Department of Forest Engineering, Düzce University, Düzce, 81620 Türkiye; b: Department of Forest Engineering, Düzce University, Düzce, 81620 Türkiye; c: Department of Forest Engineering, Bartın University, Bartın, 74100 Türkiye; d: Department of Environmental Engineering, Kastamonu University, Kastamonu,37150 Türkiye; * Corresponding author: ugurcanturk55@gmail.com
INTRODUCTION
The impact of the industrial revolution has led to the world’s population density in urban areas in the last century (Dogan et al. 2022) and, therefore, to an increase in fossil fuel consumption to meet energy demands (Yayla et al. 2022; Koc et al. 2023). Environmental pollution and global climate change occur in conjunction with each other as a result of pollution caused by industrial activities, vehicles, and other anthropogenic factors (Ghoma et al. 2023; Isinkaralar et al. 2023), and especially atmospheric pollution (Cobanoglu et al. 2023a; Key et al. 2023; Sulhan et al. 2023), urbanization (Kilicoglu et al. 2021; Istanbullu et al. 2023). Global climate change (Varol et al. 2021) has emerged as the most critical worldwide issue. These problems, especially urbanization and global climate change, are irreversible challenges that the world must confront globally (Tekin et al. 2022).
Among these problems, it is emphasized that global climate change, which is stated to affect all ecosystems, is now irreversible, and that changes related to the process should be predicted and precautions should be taken on a sectoral basis to mitigate its destructive consequences (Tekin et al. 2022; Cobanoglu et al. 2023b). Studies show that the environmental community most affected by this process is plants. Studies have also revealed that the most prominent outcomes of global climate change will manifest themselves in the form of temperature increase and drought (Koç 2022; Koç and Nzokou 2023). Because the changes will occur in a much shorter period than plants can adapt, plants that do not have adequate mobility will have difficulty adapting to the process, and individual, population, and even species losses will be inevitable (Varol et al. 2021).
Forest ecosystems are among the ecosystems that will feel the effects of the process the most because their main elements are trees with long life cycles. Therefore, for forests to survive the global climate change process with the least damage and to minimize population and species losses, it is crucial to predict possible alterations in the suitable distribution regions of trees, which are the main elements of forest areas, and take the necessary precautions. Different climate models and scenarios have been used in studies on this topic (Ardestani and Ghahfarrokhi 2021; Adhikari et al. 2022; Zhang and Wang 2023). However, to make the most accurate predictions, it is crucial to evaluate the possible results using different models and verify which models and scenarios are realized by monitoring the process and take the required protections.
Black pine (Pinus nigra Arnold.), which is one of the most common pine species, especially in the Mediterranean basin (semi-arid and arid region) in the south of Türkiye, is one of the most widespread and economically significant native conifers in Türkiye (Atalay and Efe 2010, 2012). Black pine is a drought-sensitive species that has been the subject of many scientific studies in Türkiye (Martín-Benito et al. 2008; Köse et al. 2012). It also will be affected by climate change and future projections, whose effects have been increasing over the last half-century (Andreu et al. 2007; Martin-Benito et al. 2011; Sánchez-Salguero et al. 2012). Many studies have reported that forests in the Mediterranean basin are affected by elevated temperatures and drought (Piovesan et al. 2008; Charru et al. 2017; Cetin et al. 2023). Meteorological data have revealed that the temperature in Türkiye has been on an increasing trend in the last 30 years and that the rainfall in the southern and western parts of the country has decreased in the last 70 years (Köse et al. 2017).
The current study aimed to determine the suitable distribution regions of black pine (P. nigra) in Türkiye and is of particular significance because this tree species is most frequently introduced into dryland regions, depending on different models and scenarios. The study aims are as follows:
1. To reveal the differences between these climate models using different climate models to determine the prediction accuracy of these models,
2. To compare the results of these scenarios using two different scenarios,
3. To determine the shift of suitable distribution areas in the same area in Türkiye according to different scenarios and climate models rather than the total suitable distribution area (suitable areas today versus 2060 and 2100; unsuitable areas today versus 2060 and 2100).
4. To determine the models that give the most accurate results by comparing the results of the used models with the actual distribution areas today.
Many studies have tried to determine how the suitable distribution areas of different plant species will change due to global climate change with the help of various models. Within the scope of this study, the changes in the suitable distribution areas of black pine were determined using nine different models. Thus, the aim was to determine the models with the most reliable and usable results among the models used for this purpose. The species in which the study was conducted is one of the most drought-resistant species (Topacoglu et al. 2016; Yigit et al. 2023). For this reason, the study is also crucial in determining the shift in suitable distribution areas of one of the most drought-resistant species due to the effects of global climate change.
EXPERIMENTAL
Türkiye has approximately 21 million hectares of forest areas. Within that area, 19.8% of Türkiye’s forest is covered with black pine (P. nigra Arnold.) forests (approx. 4.2 million hectares) (OGM 2015). In this study, the database of black pine was used, which has a wide distribution area within the borders in Türkiye and is of particular importance as it is used for plantation, forestation, afforestation, and recreational purposes, and is one of the species most introduced into the continental climate.
In the current research, distribution data with precisely geo-referenced coordinates were collected from publicized literature, herbarium archives, field surveys, and online sources of ‘Euforgen’ (https://www.euforgen.org/species/pinus-nigra/) and ‘Global Biodiversity Information Facility’ (https://www.gbif.org, GBIF 2023). The sampling prejudice correlated to occurrence reports (Aiello-Lammens et al. 2015) was cleared by the spatial thinning method, wherein a single existence spot was maintained in 1 × 1 km2 grid cell dimensions. For carrying out community distribution modeling of P. nigra, a total of 120 geo-referenced presence reports eventually employed spatial thinning. The distribution areas of black pine in Türkiye are presented in Fig. 1.
Fig. 1. Distribution regions of black pine in Türkiye
The authors obtained bioclimatic information from the WorldClim database version v2.1 (Fick and Hijmans 2017) (http://www.worldclim.org) at a spatial resolution of 30 arcseconds (~ 1 km). For this goal, 19 distinct bioclimatic parameters were used based on Shared Socio-economic Pathways (SSPs) 585 and 245 global climate change scenarios explained in the 2021 Intergovernmental Panel on Climate Change (IPCC) 6th Assessment Report (AR6) for the years 2060 and 2100 in the WorldClim database (Table 1). The AR6 featuring Center National de Recherches Météorologiques model version 6 [CNRM-CM6-1-HR (France)] of the IPCC with two scenarios of representative concentration pathways (SSPS 585 and SSPS 245) for two time periods (2060 and 2100) (Hamid et al. 2019) were utilized to estimate the coming possible P. nigra distribution. The current approach is among the most robust and reliable climate models used in numerous scientific research and fully mimics various climatic phenomena (Mushtaq et al. 2021; Bhat et al. 2023; Sofi et al. 2023). The authors assessed multicollinearity in bioclimatic parameters using Pearson’s correlation test to enhance model reliability and minimize inaccurate outcomes (Hamid et al. 2019; Sofi et al. 2023). A single bioclimatic parameter was chosen from a pair of associated variables with ‘r > 0.75 or < − 0.75’ correlation coefficient (Hamid et al. 2019; Rather et al. 2022; Bhat et al. 2023). The selected variables are marked with asterisks in Table 1.
Table 1. Bioclimatic Parameters Used in this Research
* = Selected variables
The niche species distribution technique is the most preferred procedure in examining the geographical distribution of plant species under climate change (Varol et al. 2021, 2022a; Zhao et al. 2021; Varol et al. 2022a; Bhat et al. 2023). Niche modeling techniques were created based on Niche theory, which is based on ecological principles and some specific algorithms related to these principles. The features of plant niches are determined by examining the existing geographical distribution of plants and various relevant environmental variable factors to predict optimum growing environments for plant species (Zhang et al. 2019; Zhao et al. 2021).
Each species distribution approach has pros and cons due to dissimilar algorithms and principles. Biomod, a modeling platform based on R software to improve the precision of predictions, was designed in 2003 and has been widely chosen in numerous studies since then (Bi et al. 2013; Zhao et al. 2021). This software package can be accessed via the CRAN web page (cran.r-project.org). It contains the subsequent species distribution modeling algorithms: surface range envelope (SRE) modeling (Busby 1991), generalized additive models (GAMs) (Guisan et al. 2002), generalized linear models (GLMs) (Guisan et al. 2002), classification tree analysis (CTA) (Vayssi’eres et al. 2000), generalized boosted models (GBMs) (Elith et al. 2008), artificial neural networks (ANNs) (Lek and Gu’egan 1999), random forest (RF) (Breiman 2001), flexible discriminant analysis (FDA) (Hastie et al. 1994), maximum entropy (MAXENT) (Phillips et al. 2006), and multivariate adaptive regression splines (MARS) (Friedman 1991).
Fig. 2. The ShinyBIOMOD workflow diagram (Obtained from https://gitlab.com/IanOndo/shinybiomod/-/blob/master/README.md)
The ShinyBIOMOD, a graphical interface for the R package “biomod2,” was operated in the present research to facilitate the SDMs (Species Distribution Models) approach (Thuiller et al. 2009, 2016). To train and apply the models, a geographical region was defined in the P. nigra distribution maps in the stand maps created by the Republic of Türkiye, General Directorate of Forestry. Afterward, the occurrence data was uploaded of P. nigra previously created and 19 bioclimatic variables. Because the dataset solely included existence reports, three pseudo-absence datasets of 120 randomly created pseudo-absence spots were created to operate the SDMs. Nine distinct SDMs (explained by Franklin 2010) were run to model P. nigra distribution. The used models were CTA, GBM, ANN, GAM, FDA, GLM, RF, MAXENT, and MARS. This modeling approach was made using the instructions defined at ‘https://gitlab.com/IanOndo/shinybiomod/-/blob/master/ README.md’, and the workflow chart is exhibited in Fig. 2. Some studies have recently given a detailed explanation of this model (Henderson et al. 2022; Kass et al. 2023; Sillero et al. 2023).
Pseudo-absence and occurrence spots were randomly split into verification (20%) and training data (80%). To specify model statistics and P. nigra species distribution, three replicates were run for all models on each of three pseudo-absence datasets (9 total replicates for each model). Area Under the Curve (AUC) error evaluation data, Receiver Operating Characteristics (ROC), and True Skill Statistic (TSS) were used to evaluate the performances of each model recurrence (Allouche et al. 2006; Fawcett 2006; Canturk and Kulaç 2021; Varol et al. 2021). Lastly, ensemble models were built to estimate P. nigra distribution across all nine modeling approaches. The authors designed three distinctive projections of P. nigra distribution operating ensemble modeling outputs. These outcomes were estimated binary predictions of absence/presence, likelihood of occurrence, and forecasted probability of presence by committee agreement of binary separated model outcomes. Binary projections were built by generating a threshold of calculated possibility to maximize the perceptiveness and accuracy of the model. This situation indicates that a threshold (TSS and ROC ≥ 0.8) was used for probability estimations to the accurate negative ratio (specificity) and maximize the accurate positive ratio (sensitivity) through the whole range. Pixels with probability weights above the threshold were classed as presences, and pixels below the threshold were classified as absences. The current modeling method was employed and described in detail in a paper on American chestnuts by Henderson et al. (2022). The current procedure lets the client specify and modify a model by changing the initial situations, classifications, model parameters, and boundary circumstances to acquire the most reasonable prediction outcomes (Luo et al. 2017).
In the current study, the existing suitable distribution territories of the species were first distinguished according to the models used. In the next stage, the suitable distribution regions of the species in 2060 and 2100 were modeled according to SSPs 245 and SSPs 585 scenarios. Then, the proportional changes in 2060 and 2100 according to the suitable distribution areas today were calculated and evaluated according to different models.
RESULTS
Accuracy statistics for each distribution approach are shown in Table 2. These illustrate the average values for ROC and TSS across all model repetitions. The ROC values expand from 0 to 1 and represent the likelihood that assigning a forecasted fitness weight at a random spot of presence is higher than setting a predicted fitness weight at a random dot of absence (Fawcett 2006). The TSS expands from −1 to 1, with 0 denoting an approach that acts no better than random guesses (Fawcett 2006; Henderson et al. 2022). The AUC values of 0.6 and 0.7 are commonly decoded as “fair” models, and ROC weights between 0.7 and 0.8 are meant as “good” models.
Table 2. Comparison of Pinus nigra Distribution Models’ Performance Based on Each Model’s Validation Data against the Test Data
Suitable distribution areas of P. nigra throughout Türkiye were determined according to different models, and suitable distribution area maps are given in Fig. 3, and numerical values of distribution areas are given in Table 3.
Fig. 3. Current suitable distribution areas of Pinus nigra
It can be seen that the suitable distribution regions for P. nigra vary considerably according to different models, and, generally, the northern and southwestern parts of the central part of Türkiye are suitable regions for the distribution of the species in all models (Fig. 3). The models that offer the broadest suitable distribution area for the species are MAXENT (37.58%), GLM (34.27%), and GAM (29.33%), while the models that offer the narrowest distribution area are RF (7.71%), FDA (14.79%), and GBM (15.47%) (Fig. 3 and Table 3). Regarding the SSPs 245 scenario, the suitable distribution regions of the P. nigra in 2060 are given in Fig. 4, and the numerical data are shown in Table 4.
Table 3. Current Areal Distribution of Suitable Distribution Areas of Pinus nigra
Fig. 4. Suitable distribution areas of Pinus nigra in 2060 based on the SSPs 245 scenario
Table 4. Suitable Distribution Areas of Pinus nigra in 2060 Based on the SSPs 245
When Fig. 4 and Table 4 are examined, it is predicted that there will be a slight increase in the suitable distribution regions of P. nigra until 2060, only according to the GAM (1.16%), CTA (0.51%), and GBM (0.21%) models. According to all other models, the suitable distribution regions of P. nigra will decrease compared to today, and this decrease may reach 3.39%, 2.51%, and 2% in the total area according to MAXENT, FDA, and MARS, respectively. Considering the suitable distribution area, it is predicted that the suitable distribution regions of P. nigra may decrease nearly 17% (according to FDA) in 2060, according to the SSPs 245 scenario. Regarding the SSPs 245, the suitable distribution regions of the P. nigra in 2100 are given in Fig. 5, and its numerical data are given in Table 5.
Fig. 5. Suitable distribution areas of Pinus nigra in 2100 based on the SSPs 245 scenario
Table 5. Suitable Distribution Areas of Pinus nigra in 2100 Based on the SSPs 245 Scenario
Regarding the SSPs 245 scenario, the suitable distribution regions of P. nigra will increase in 2100, unlike 2060, according to most of the models subject to the study (Fig. 5 and Table 5). This increase will be seen primarily in the FDA, MARS, and MAXENT models regarding total area. According to the CTA model, the suitable distribution regions of the P. nigra will decrease significantly compared to 2060, and there will be a slight decrease according to the ANN and GLM models. It was defined how the appropriate distribution regions of the species will change between now and 2100. This change is shown on the map in Fig. 6, and its numerical data is given in Table 6.