Assessing climate-driven shifts in Liquidambar orientalis using ensemble species distribution models in the Eastern Mediterranean Region Authors: Derya Evrim Koç, Demet Biltekin, Beyza Ustaoğlu Published: 2025 ABSTRACT AbstractBackgroundLiquidambar orientalis, a relict and endemic tree species of the Eastern Mediterranean, is increasingly threatened due to its narrow distribution and intense anthropogenic pressures. This study aims to model the spatiotemporal changes in its habitat suitability from the Last Glacial Maximum (LGM) to future climate scenarios (2050 and 2070) using a robust ensemble species distribution modeling (SDM) framework. By integrating paleobotanical data with predictive modeling, we provide crucial insights into the species' ecological resilience and future conservation priorities.ResultsThe ensemble models showed excellent predictive performance (AUC = 0.96, TSS = 0.91, Boyce Index = 0.84, Kappa = 0.89), identifying mean diurnal range (Bio2) and precipitation of coldest quarter (Bio19) as the most influential variables. Fossil pollen records confirm the species long–term persistence in southwestern Anatolia since the early Miocene. While projections under the RCP 2.6 scenario suggest potential habitat expansion or stability, especially in mountainous refugia, the RCP 8.5 scenario indicates a reduction in highly suitable areas. Nevertheless, key ecological niches are expected to persist, particularly along the southern slopes of the Taurus Mountains.ConclusionsThis is the first study to combine multi temporal ensemble SDMs and fossil records for L. orientalis, revealing its ecological flexibility over millennia. The findings underscore the necessity of prioritizing genetically diverse populations and climatically stable refugia in conservation strategies. Our integrative approach provides a valuable framework for assessing the climate resilience of other Mediterranean relict species.Keywords: Ensemble Modelling, Liquidambar orientalis, Climate change, Biodiversity, Endangered species, Pollen grains Introduction The impact of climate on ecosystems and biodiversity is a significant factor that must be considered in global environmental studies. In particular, climate change has emerged as a significant factor influencing the composition and distribution of natural habitats and species, with shifts in temperature and precipitation patterns [1–3]. Increases in greenhouse gas emissions, rising global average temperatures, rising sea levels and the frequency of climatic events present significant threats to ecosystem balance. Biodiversity is a fundamental component of the healthy functioning of ecosystems. However, the effects of climate change may result in a narrowed range of habitats for many species, leading to population decline and, in some cases, local or global extinction [4, 5]. This has the potential to impair the functionality of ecosystems. Quaternary climatic shifts significantly shaped current vegetation, though Tertiary events also played a role. High temperatures and humidity in the Tertiary influenced plant habitat expansion, contraction, and extinction [6]. Similarly, during the Quaternary's Last Glacial Maximum (LGM), falling temperatures altered plant distribution, evolution, and diversity [7–10]. Future climate change is predicted to cause plant species to migrate to different altitudes or latitudes, with some facing extinction as they move out of their climatic niches [11, 12]. Türkiye, situated at the crossroads of the Mediterranean, Caucasian, and Irano–Anatolian biodiversity hotspots, boasts exceptional plant diversity and endemism [13]. Species distribution models (SDMs) are fundamental tools for elucidating the impact of climate change on biodiversity, formulating conservation and management strategies, and making spatial and temporal predictions [14–16]. They also facilitate temporal and spatial predictions. In recent years, there has been a notable increase in the number of studies conducted with SDMs [15–18]. Nevertheless, the proliferation of modelling algorithms has rendered the selection of the most suitable methodology a challenging endeavour. Biomod, which analyses species distributions using community modelling, provides significant advantages in resolving these problems by eliminating methodological uncertainties and investigating species–environment relationships [19–21]. Predicting the future distribution of especially endemic and relic tree species with limited distribution and investigating their adaptation mechanisms to global climate change are of great importance for ensuring their sustainability and numerous studies have been conducted on this subject [22–29]. However, although there are studies on this subject, none of them have applied ensemble modeling approaches to Liquidambar orientalis. Previous studies have provided valuable insights into the potential impacts of climate change on the distribution of Liquidambar orientalis. Ürker et al. (2023) emphasized that rising temperatures could create opportunities for the recovery of this endangered species, but their analysis was restricted to future projections and did not incorporate paleoecological evidence [24].Similarly, Kenar (2024) modeled the species’ range shifts across the Mid–Holocene, present, and 2070, identifying precipitation of the wettest quarter as the key climatic driver; however, the study considered a narrow set of time periods, a single modeling approach, and lacked integration with fossil data [28]. Building upon these earlier efforts, the present study introduces a broader temporal framework, encompassing the Last Glacial Maximum (LGM), Mid–Holocene, present, and two future horizons (2050 and 2070) under both RCP 2.6 and RCP 8.5 scenarios Fossil pollen records were therefore used only as historical context, not as temporal validation of the model projections, and the results were interpreted with appropriate caution given the difference in temporal scale between the Miocene pollen evidence and the late Quaternary periods modelled in this study. The genus Liquidambar has existed for approximately 65 million years [30]. L. orientalis, or Anatolian sweetgum, is an important relict endemic species native to the Eastern Mediterranean, persisting from the Tertiary period [31–34]. This rare European tree [35] belongs to the Altingiaceae family (formerly Hamamelidaceae) [36]. Glaciations restricted Liquidambar species to refugial areas in Türkiye, eastern America, and eastern Asia, with Anatolia serving as a significant refuge [33, 36]. Its Anatolian distribution is influenced by the Mediterranean climate, favoring mild, wet winters and hot, dry summers, and thriving near rivers in well-drained soils [37]. Historically, Liquidambar species were used in medicine and perfumery [38–44]. Today, Anatolian sweetgum oil is valuable in medical, chemical, and cosmetic industries [45]. It's also used for landscaping. Classified as Endangered (EN) by IUCN and recognized for protection by EUFORGEN in 2001 [32, 46], understanding its distribution changes from the Last Glacial Maximum (LGM) to the present and future is crucial. This data can inform conservation policies to protect the genus from future climate impacts. Climate models can also offer pertinent information at broader spatial and temporal scales. To date, while some studies exist on the distribution of this genus, none have employed ensemble modeling approaches spanning from the Last Glacial Maximum (LGM) to the present. In this study, ensemble models were used to model the alterations in the Liquidambar distribution in response to climatic conditions during the LGM, the Middle Holocene (MH), the present (1960–2000), and future projections for 2050 (average for 2041 − 2060) and 2070 (average for 2061 − 2080). Our objectives include the following questions: 1) What are the primary climate factors governing Liquidambar distribution? 2) What is the characterization of distribution changes from the LGM to the present? 3) How will the Liquidambar distribution evolve in response to future climate warming scenarios? Materials and methods Features of Liquidambar orientalis Mill. The genus Liquidambar has four species in the Altingiaceae family, ranging from North America to East Asia (Fig. 1) [47, 48]. L. orientalis is the only endemic species in the Mediterranean Basin and a Tertiary relict, with two varieties (L. orientalis var. orientalis and L. orientalis var. integriloba) showing minimal genetic differentiation due to hybridization [49]. It is listed as "Endangered (EN)" on the IUCN Red List [46]. Distribution areas of all species of Liquidambar and fossil pollen and occurrence data distribution map for L. orientalis. The numbers refer to fossil pollen localities: 1) Ermenek (Karaman), 2) Güvem Basin (Ankara), 3) Tavşanlı (Kütahya), 4) Çan (Çanakkale), 5) Soma Basin (Manisa), 6) Çatakbağyaka (Muğla), 7) Burhanlı (Çanakkale), 8) Bayramhacılı (Nevşehir), 9) İntepe (Çanakkale), 10) Enez (Edirne), 11) Avadan (Mersin), 12) DSDP site 380 A (Black Sea), and 13) Lake Köyceğiz (Muğla) Liquidambar orientalis, which can reach a height of approximately 20 m, possesses dark ash brown, fissured bark [50]. This monoecious species undergoes its period of decay between March and April. Fruiting heads typically begins in April and continues through May. Seed shedding occurs between August and October [51, 52]. L. orientalis is generally characterised by a slow growth rate and a high capacity for producing root and stump shoots. However, its root system is shallow, exhibiting superficial development [53]. The geographical distribution of the Anatolian sweetgum (Liquidambar orientalis) ranges from sea level up to an altitude of 500 m, although some individuals have been reported at elevations as high as 1000 m. This species prefers an annual precipitation range of 1000–1200 mm and an average annual temperature of approximately 18 °C. It can grow in sloping and arid soils; however, optimal growth occurs in nutrient rich, deep, and moist soils, particularly along riverbanks and coastal areas [52, 54]. Its distribution in southwestern Türkiye includes Muğla, Denizli, Burdur, Isparta, and Antalya, along with Rhodes Island [31, 37, 55, 56]. While found inland as small islands, it thrives in humid–warm bioclimatic zones, especially in shallow swamps around Lake Köyceğiz, Muğla (Fig. 1). It forms extensive forests in high water table areas and gallery forests along stream banks, typically occurring up to 500 m above sea level, though some can reach 1000 m. Annual rainfall in these regions ranges from 900 to 1250 mm. Data collection Occurrence data Occurrence data for Liquidambar orientalis were compiled from the GBIF [57] and EUFORGEN [58] databases, as well as from the literature [55, 56], to identify the species natural distribution areas. Initially, approximately 130 records were collected. All database and literature records were carefully examined, and erroneous or duplicate entries were removed. Environmental variables were obtained from WorldClim database at a spatial resolution of 2.5 (~ 5 km). To ensure consistency between the occurrence data and the environmental grid, a pixel based thinning procedure was applied, retaining a single record per raster cell [59–61]. Because all 87 L. orientalis occurrence points were already located in unique 5 km cells, this filtering step did not reduce the sample size but confirmed that each record represented a distinct set of environmental conditions. As a result, a total of 87 reliable occurrence records were obtained and included in the modeling. The occurrence dataset used for model calibration included all known natural populations of Liquidambar orientalis in southwestern Turkey. In addition, three in situ conservation sites registered in the EUFGIS database were incorporated: TUR00200 (Aydın) [62], TUR00037 (Marmaris–Gökova) [63] and TUR00109 (Isparta–Bucak) [64]. These locations represent the core genetic conservation units of the species and correspond to the central part of its natural distribution range. Field studies were conducted in June 2019 and June 2022 in Muğla Province (Köyceğiz, Turgut, Adaköy, Aksaz, Reşadiye Peninsula) with the objectives of identifying distribution zones, collecting plant samples, and recording geographic data. The surveyed sites are located along river valleys, which host partially dense and accessible populations of the species. The collection of specimens and subsequent photographing in their natural habitats was undertaken with the objective of supporting species identification. The identification of Liquidambar orientalis was carried out in situ by Derya Evrim Koç, who made a comprehensive analysis of the species distinct morphological characteristics, including leaf shape, bark texture, and habitat specificity. However, voucher specimens were not deposited in a public herbarium (Fig. 2). Photographs of L. orientalis Miller from Marmaris (Muğla, southwestern Türkiye, photo by Derya Evrim Koç): A general habitat near streams; B branches of Turkish sweetgum on the trunk; C and D) leaves of L. orientalis Climate data This study focuses on modelling the ecological niche of L. orientalis primarily via climatic variables, given their significant influence on plant distributions. We utilized ecological niche modelling to predict species distributions in response to the climate conditions of the Last Glacial Maximum (LGM), Middle Holocene (MH), and future climate scenarios (RCP 2.6 and 8.5), employing data from the Community Climate System Model (CCSM4). This model has been widely used in similar studies in Anatolia [64–70]. For the climatic data, information was obtained from the WorldClim database [71], which uses a spatial resolution of 2.5 min (~ 5 km) covering the period from 1960 − 1990. It is acknowledged that multiple climate datasets are available, including WorldClim 2.1 and CHELSA. However, it should be noted that WorldClim 2.1 datasets do not provide the necessary paleoclimatic data for earlier periods. Although CHELSA provides high-resolution climate surfaces and continuous paleoclimate coverage, its LGM layers require additional preprocessing steps, including temporal averaging and manual masking of marine areas. These steps may introduce methodological subjectivity and complicate reproducibility across modelling workflows In order to ensure full comparability and transparency, we preferred WorldClim 1.4, which offers standardized and widely used CCSM4 LGM and mid-Holocene layers. WorldClim remains the most established paleoclimate dataset in SDM studies, particularly where reproducibility and methodological consistency are prioritized. This choice also ensures the comparability of the study with previous research that has used the same dataset, thereby maintaining methodological consistency and supporting robust interpretation of the results. Understanding the interplay between species and environmental variables is essential for accurately predicting current and future species distributions. In species distribution models, environmental layers derived from raw data serve as predictor variables and can highlight spatial autocorrelation associated with a species geographic range [72–75]. To prevent potential biases from multicollinearity among correlated bioclimatic variables, we conducted Pearson correlation analysis on 19 bioclimatic layers using the SDMtoolbox 2.5 [76] in ArcGIS Pro. Variables exhibiting a correlation coefficient greater than 0.85 were excluded, resulting in a final set of 11 variables used for modelling: Bio1 (annual mean temperature), Bio 2 (mean diurnal range), Bio3 (isothermality), Bio4 (temperature seasonality), Bio8 (mean temperature of wettest quarter), Bio9 (mean temperature of driest quarter), Bio12 (annual precipitation), Bio13 (precipitation of wettest month), Bio14 (precipitation of driest month), Bio15 (precipitation seasonality), and Bio19 (precipitation of coldest quarter) (Table 1). Additionally, species‐specific ecological relevance was considered when reducing collinearity; therefore, Bio1 was retained in the final set, as annual mean temperature is a key ecological constraint for L. orientalis. Table 1Correlation matrix between 19 bioclimatic variablesBio1Bio2Bio3Bio4Bio5Bio6Bio7Bio8Bio9Bio10Bio11Bio12Bio13Bio14Bio15Bio16Bio17Bio18Bio19Bio11.00Bio20.321.00Bio30.340.521.00Bio40.110.56−0.401.00Bio50.910.600.250.491.00Bio60.930.030.39−0.240.701.00Bio70.090.77−0.140.950.49−0.281.00Bio80.19−0.050.01−0.050.110.18−0.081.00Bio90.790.370.250.230.810.700.23−0.171.00Bio100.970.430.210.360.980.810.320.160.801.00Bio110.960.140.44−0.180.760.99−0.190.190.720.861.00Bio12−0.30−0.62−0.24−0.40−0.47−0.11−0.50−0.05−0.40−0.38−0.171.00Bio13−0.05−0.47−0.10−0.34−0.220.12−0.43−0.07−0.16−0.120.070.851.00Bio14−0.47−0.52−0.20−0.40−0.62−0.30−0.460.23−0.66−0.54−0.350.620.221.00Bio150.690.470.290.310.750.550.33−0.100.650.730.60−0.290.16−0.781.00Bio16−0.04−0.46−0.12−0.32−0.200.12−0.41−0.08−0.15−0.110.070.870.990.230.161.00Bio17−0.47−0.54−0.19−0.44−0.63−0.29−0.490.22−0.64−0.55−0.340.650.240.99−0.790.251.00Bio18−0.54−0.43−0.15−0.35−0.65−0.40−0.390.30−0.80−0.60−0.440.610.270.87−0.660.280.871.00Bio190.29−0.320.04−0.290.120.43−0.37−0.250.290.210.390.660.84−0.030.340.850.00−0.131.00 Correlation matrix between 19 bioclimatic variables Individual and ensemble modelling Species distribution models are essential for identifying habitats and improving conservation efforts [25, 77–79]. While tools like GARP, ENFA, Bioclim, and Maxent are widely used [80–83], ensemble modeling offers greater reliability by combining multiple models, reducing inconsistencies, and yielding more precise analyses [19, 84, 85]. R is widely used in biodiversity research, with existing packages focusing on species number and abundance [85]. Biomod, the first R package for Species Distribution Models (ENMs), was published in 2009 by Thuiller et al. It allows for modeling and testing species distributions under varying environmental conditions using diverse techniques [86]. We used biomod2 (version 4.2–5.2) in R (version 4.2–5.2) for ensemble distribution modeling of Liquidambar orientalis. Biomod2 applies 8 algorithms for species–environment relationships: Classification Tree Analysis (CTA) [87], Flexible Discriminant Analysis (FDA) [88], Generalized Additive Model (GAM) [89], Generalized Linear Model (GLM) [90], Generalized Boosted Model (GBM) [91], Multiple Adaptive Regression Splines (MARS) [92], Maximum Entropy (MAXENT) [93], Random Forest (RF) [94], Random Forest downsampled (RFd), and eXtreme Gradient Boosting (XGBoost) [95]. The models were calibrated using the settings included in the ‘bigboss’ strategy, as recommended by the developers of the biomod2 package, Pseudo–absence points were generated randomly within the boundaries of the study area. While most algorithms utilized presence–absence data, MaxEnt does not rely on pseudo–absence data; instead, it uses background points sampled from the study area. For model training and evaluation, 80% of the data were used for training and 20% for testing. Additionally, ten distinct pseudo-absence datasets were generated for each algorithm, each containing 1000 pseudo absence points. This number was selected because using 1000 pseudo absences improves model stability and reduces prevalence imbalance, as recommended by simulations [96]. This number is also consistent with recent SDM applications employing Biomod2 for tree species [97, 98]. Models were subjected to ten repetitions of cross validation (ten evaluations). Model accuracy evaluation Due to potential misleading outcomes from single metrics [99–102], we used multiple evaluation criteria for species distribution models: AUC, TSS, Kappa, and the Boyce Index [103–105]. AUC assesses model success across all thresholds; values above 0.75 indicate good performance, with higher values approaching 1 signifying increased accuracy [106–108]. TSS considers false negatives/positives and random guesses, with values near + 1 indicating excellent performance [93]. The Boyce Index, a reliable measure for presence–based predictions, ranges from –1 to 1; positive values show accurate predictions, near zero indicates random chance, and negative values suggest incorrect predictions of low quality areas [25, 106]. For modeling, a median ensemble based on TSS was used [79]. Continuous maps were reclassified into binary maps using the sensitivity–specificity approach, which measures omission (sensitivity) and commission (specificity) errors [107–109]. Ensemble models were developed for present and future periods using committee average and weighted mean approaches [108]. A permutation approach determined the relative impact of climatic variables on Liquidambar orientalis [14]. Variable importance was calculated by comparing original and permuted estimates, with values near 1 indicating high contribution. Model results were imported into ArcGIS Pro for reclassification and raster calculations to visualize and quantify changes in Liquidambar orientalis distribution under current and future climates. The reclassification threshold was determined by maximizing the sum of training sensitivity and specificity and optimizing TSS [109, 110]. Regions with a logistic output of p ≥ P were deemed suitable; otherwise, they were unsuitable [95, 111–114]. Suitable areas were categorized into low (P ≤ p < 0.5), medium (0.5 ≤ p < 0.7), and high (p ≥ 0.7) suitability. Suitability change analysis, using a four class system (species loss, absence, stable, and gain), provides insights into habitat changes [115, 116]. Chronology of the fossil data The considered (eight outcrops, two cores: marine and lake) sections are dated via different dating methods (Table 2). The Ermenek lacustrine sediments were dated to the lower Miocene via the analysis of mammalian fossils and freshwater ostracods. In the Güvem Basin, sediments comprising lacustrine diatomites and clays at altitudes of approximately 1500 m are dated by Burdigalian mammal fossils, which are discovered within volcanic deposits that range in age from 20 to 18 million years [117]. The fluvial–lacustrine deposits at Çatakbağyaka, located at 500 m, are dated to the Langhian based on mammal remains [118, 119]. In the higher Nevşehir region, clays at Bayramhacılı (1000 m) were dated through the radiometric assessment of associated ignimbrites, with an estimated age range of 7.6 − 6.8 million years [120]. Additionally, low–pollen floras from Burhanlı and İntepe were analysed via calcareous nannoplankton fossils dating from the late Tortonian to the earliest Zanclean [121]. Marine shales in the Avadan section were dated to the early Zanclean stage via planktonic foraminifera [122]. DSDP Site 380, which is located at a depth of 2107 m, uses calcareous nannoplankton biostratigraphy to achieve high resolution climatostratigraphy, indicating confident dating for the past 6 million years [33, 123, 124]. This study also reports the distribution of Turkish sweetgum trees in the Soma Basin and Lake Köyceğiz, which is supported by radiocarbon ages and age modelling [125–127]. Key mammalian fossils on the southwestern side of the Soma Basin suggest a late early Miocene age (MN3 − MN4) [128]. Moreover, 40Ar/39Ar dating of volcanic rocks indicated early Miocene development in the Soma Basin, likely predating the Burdigalian [129]. Early Miocene sediments in Tavşanlı and Çan were dated via radiometric K–Ar dating and macroflora analysis [130]. Finally, nanofossils from Enez Lagoon were identified as late Messinian to early Zanclean [118]. Table 2Fossil data of LiquidambarNumberLocalitiesRegionAgesMaterialReferences1Ermenek (Karaman)Southern TürkiyeEarly Miocene (Aquitanian)Mammals and freshwaterostracods [131, 132]2Güvem Basin (Ankara)Central TürkiyeEarly Miocene (Burdugalian)Mammals and volcanic deposits [117]3Tavşanlı (Kütahya,Western TürkiyeEarly MioceneRadiometric (K/Ar dating) [130, 133–136]4Çan (Çanakkale)Western TürkiyeEarly MioceneRadiometric (K/Ar dating) [130, 133–136]5Soma Basin (Manisa)Western TürkiyeEarly MioceneMammals and Radiometric (K/Ar dating) [126, 128, 129]6Çatakbağyaka (Muğla)South-western TürkiyeMiddle Miocene (Langhian)Mammals [118, 119, 137, 138]7Burhanlı (Çanakkale)North-West TürkiyeLateTortonianCalcareous nannoplankton [121, 124]8Bayramhacıl(Nevşehir)Central TürkiyeLate MioceneRadiometry of associated ignimbrites [120]9İntepe (Çanakkale)North-West TürkiyeLateMessinian to earliest ZancleanCalcareous nannoplankton [121, 139]10Enez (Edirne)Nort-western TürkiyeLate Messinian-early ZancleanCalcareous nannoplankton [121, 124]11Avadan (Mersin)Southern TürkiyeEarly PliocenePlanktonicforaminifers [122]12DSDP Site 380ABlack SeaLate Miocene to PresentCalcareous nannoplankton [33, 123, 124]13Lake Köyceğiz (Muğla)South-western TürkiyeLate HolocenePeat and gyttja [125, 127] Fossil data of Liquidambar Mammals and freshwater ostracods Late Tortonian Late Messinian to earliest Zanclean Planktonic foraminifers Results Model evaluation The accuracy of the produced ensemble models was as follows: an AUC value of 0.96, a TSS value of 0.92, a Kappa value of 0.92, and a Boyce Index of 0.84 for committee averaging; and an AUC value of 0.96, a TSS value of 0.91, a Kappa value of 0.89, and a Boyce Index of 0.93 for the weighted mean. The AUC, TSS, Kappa and Boyce Index values indicate that the final models are good at predicting Liquidambar orientalis. Artificial Neural Networks (ANN) and Surface Range Envelope (SRE) ensemble models, which have low model accuracy values, are not preferred in ensemble modeling because they will negatively affect usability performance (Fig. 3). The elevated area under the curve (AUC), true skill score (TSS), Kappa, and Boyce Index values for both ensemble models (committee averaging and weighted mean) show that they performed well in predicting suitable habitat areas for Liquidambar orientalis. This finding suggests that the models effectively captured the relationships between environmental factors and the presence of the species (Fig. 4). Workflow diagram of the data and processes used to develop ensemble modeling of suitable habitats for L. orientalis The area under the curve (AUC) of the receiver operating characteristic (ROC) curve, The true skills statistics (TSS), Boyce Index and Kappa indicate the model evaluation scores for individual and ensemble models Bioclimatic variable contributions The relative contributions of the bioclimatic variables varied across the different modeling algorithms Following comprehensive evaluation of all available algorithms, it was determined that the variables contributing most significantly to the eleven selected bioclimatic variables were Bio2 (0.46), Bio19 (0.44), Bio4 (0.34) (Table 3). Table 3Algorithm–wise values of the bioclimatic variables used for Liquidambar orientalis modelling (GLM: Generalized Linear Model, GBM: Generalized Boosted Model, GAM: Generalized Additive Model, CTA: Classification Tree Analysis, FDA: Flexible Discriminant Analysis, RF: Random Forest, RFd: Random Forest down sampled, MARS: Multivariate Adaptive Regression Splines, MaxEnt: Maximum Entropy, XGBoost: eXtreme Gradient BoostingBioclimatic variablesIndividual species distribution modelsGLMGBMGAMCTAFDARFRFdMARSMaxEntXGBoostMeanBio10.170.010.000.000.000.010.000.000.000.000.01Bio20.790.300.880.440.790.110.000.860.170.350.46Bio30.310.020.740.000.050.030.010.570.030.030.17Bio40.840.011.000.000.050.030.040.970.490.000.34Bio80.600.010.100.000.000.010.000.000.020.000.07Bio90.600.030.560.000.000.020.010.000.000.000.12Bio120.670.040.470.000.070.050.010.020.000.000.13Bio130.700.371.000.001.000.150.040.980.120.040.44Bio140.390.000.230.000.000.020.000.000.000.000.06Bio150.710.080.160.000.000.070.040.000.200.030.12Bio190.970.590.020.960.180.150.120.290.360.820.44 Algorithm–wise values of the bioclimatic variables used for Liquidambar orientalis modelling (GLM: Generalized Linear Model, GBM: Generalized Boosted Model, GAM: Generalized Additive Model, CTA: Classification Tree Analysis, FDA: Flexible Discriminant Analysis, RF: Random Forest, RFd: Random Forest down sampled, MARS: Multivariate Adaptive Regression Splines, MaxEnt: Maximum Entropy, XGBoost: eXtreme Gradient Boosting Given that Anatolian sweetgum thrives in regions characterized by an average annual temperature of approximately 18 °C and annual average precipitation ranging from 1000–1200 mm [32], its growth is influenced by factors such as temperature, precipitation requirements, and paleoclimatic changes. Consequently, it is reasonable for the models to prioritize Bio2 (mean diurnal range) and Bio19 (precipitation of coldest quarter) as the most critical variable (Table 3). Fossil records of L. orientalis and comparison of the past predicted results Fossil pollen reveals Liquidambar existed in Anatolia since the early Miocene (approx. 23 million years ago), initially distributed across southern, southwestern, and central regions [33, 117, 126, 130–132]. By the middle Miocene, it expanded to northwestern Anatolia [124]. Pollen records confirm its presence in northwest Türkiye from the late Miocene to early Pliocene [33, 124, 139], and regularly in Black Sea cores until 1.28 million years ago. However, Liquidambar disappeared from north–northwest Anatolia around 270,000 years ago [33]. The integration of multiple data sources, including fossil records and species distribution modelling (SDM), provides a comprehensive understanding of the historical and potential habitats of L. orientalis. The analysis from the Last Glacial Maximum (LGM) model indicates that suitable habitat regions for L. orientalis exist along the Aegean and Mediterranean coasts, specifically in areas such as Aydın, Muğla, Antalya, and Mersin in southern and southwestern Türkiye (Figs. 5 and 6). Additionally, favourable ecological conditions for the species are likely found beyond Anatolia, extending to regions such as Rhodes Island, Cyprus, and Lebanon. As the climate transitioned from the LGM to the middle Holocene (MH), significant shifts in species distribution were observed. Notably, highly suitable habitats within the Reşadiye Peninsula (Datça Peninsula) and its northern areas experienced a decline. Conversely, regions around Köyceğiz have sustained suitable habitat for L. orientalis, where community formations continue to be robust. It is anticipated that highly suitable zones may still exist, particularly in more localized areas along river valleys and adjacent landscapes. Predicted habitat suitability for Liquidambar orientalis under the LGM, MH, current and future conditions based on the committee average Predicted habitat suitability for Liquidambar orientalis under the LGM, MH, current and future conditions based on the weighted mean The current distribution model of L. orientalis closely resembled that of the MH; however, the potential habitat areas around Antalya experienced a degree of reduction, as depicted in (Figs. 5 and 6). Quantitatively, during the LGM, unsuitable habitat constituted 96.6% for ca approach and 95.2% for wm approach of the overall area, with medium suitable and highly suitable habitats accounting for 0.5% (ca), 0.7% (wm) and 2.0% (ca), 0.3% (wm) respectively. In the MH, the proportions shifted slightly, with unsuitable habitats increasing to 94.6% (ca), 95.8% (wm); medium suitable habitats making up 2.0% (ca), 1.1% (wm), and high suitable habitats declining to 1.0%(ca), 1.1% (wm) (see, Figs. 7 and 8). In Lake Köyceğiz, Liquidambar is best represented in the upper part of the pollen record, suggesting that in more recent times, suitable habitats for this sweetgum expanded [125]. Potential habitat areas of L. orientalis in the LGM and MH, current and future based on the committee average Potential habitat areas of L. orientalis in the LGM, MH, current and future based on the weighted mean Assessment of the present and predicted distributions The current potential distribution patterns of L. orientalis are illustrated in (Figs. 5 and 6). The analysis revealed that potentially suitable habitats for the species are predominantly concentrated in mountainous regions along the Mediterranean coast of Türkiye. Notably, areas of high suitability not only align closely with the actual observed distribution sites in southwestern Türkiye but also extend beyond these observed areas, indicating a greater potential range for L. orientalis. In the Mediterranean region, high suitable habitats account for ca approach approximately 1.1% and for wm approach 0.7% of the total area, as detailed in Figs. 7and 8. This high suitable area is located primarily in key geographic features, such as the western regions of the Anatolian Plateau, the southern Taurus Mountains, and the anti–Taurus Mountains. In addition to high suitable areas, medium suitable habitats encompass an area of 1.7% for ca approach and 2.3% for wx approach. These medium suitable zones are primarily found in regions such as the western Anatolian Plateau, western Taurus Mountains, and parts of the Taurus and anti–Taurus belts. Low suitability habitats cover a relatively large spatial extent of 1.0% for ca approach and 1.9% for ewm approach are situated mainly on the Anatolian Plateau. This suggests that while there are regions where L. orientalis may struggle to thrive, there is still a notable presence of potentially suitable conditions. Conversely, potentially unsuitable areas constitute the majority of the landscape, covering a substantial over 90% of the total area. These unsuitable regions are represented in the analysis as white areas in (Figs. 5 and 6), highlighting the limited distribution potential for L. orientalis outside of the identified suitable habitats. Suitable areas for L. orientalis in the future The results indicate a pronounced concentration of suitable habitats for Liquidambar orientalis in the mountainous terrains along the Mediterranean coast, with broader areas of potential distribution indicated by the medium and low suitability classifications, while predominantly unsuitable habitats occupy the vast majority of the region (Figs. 5 and 6). The potential distribution of sweetgum trees under two climate change scenarios, RCP 2.6 and RCP 8.5, was projected for the year 2050. Overall, the model predictions indicate a significant degree of continuity with current suitable habitat areas, suggesting that the ecological conditions for the species may not change drastically in the near future. This continuity can be attributed to expectations that climate impacts in 2050 will largely reflect existing conditions, particularly in regions where the species is currently thriving. According to the model projections, L. orientalis is likely to maintain its current ecological conditions and even expand into new suitable distribution areas, particularly in sheltered regions, such as the southern slopes of the Amanos Mountains near Antalya in southern Türkiye. The anticipated distribution maps for 2050 and 2070 (Figs. 5 and 6) illustrate the expected changes in suitable habitat across three distinct concentration pathways, namely, RCP 2.6 and RCP 8.5. Within the Mediterranean region, the analysis revealed a declining trend in regions classified as having high and medium suitability under both RCP scenarios. Notably, the reduction in high suitability areas was more pronounced under the RCP 8.5 scenario, where the extent of areas deemed highly suitable reached its minimum values by 2050 (Figs. 5 and 6). Conversely, the low suitability areas continued to decrease, indicating a shift in habitat conditions favouring less suitable landscapes. The extent of unsuitable habitats increased notably under RCP 8.5. Particularly in western Türkiye, the projection indicates a significant reduction in regions deemed suitable for L. orientalis. Under the RCP 8.5 scenario for 2050, high suitability areas diminished, whereas medium suitability areas markedly increased under RCP 2.6. The models indicate an overall decrease in low suitability areas, with a minor increase in unsuitable habitats. This suggests a complex interplay of environmental factors that could increase the number of medium suitability regions, even as high suitability areas decline. A comparison of the conditions expected in 2050 to those in 2070 highlights a concerning trend, with habitat areas projected to increase under both RCP scenarios. For RCP 2.6, medium suitability areas are expected to grow, in contrast with a decrease in those areas under RCP 8.5. In terms of low suitability habitats, an increasing trend was noted across all RCPs except for RCP 8.5, where conditions remained stable. Currently, 99.0% (ca), 98.9% (wm) of the habitat area is classified as unsuitable, for both approach 0.04%) identified as suitable (Fig. 7). This distribution underscores the precarious status of habitats suitable for L. orientalis. The relative stability in habitat conditions from the Last Glacial Maximum (LGM) to the present is indicative of the adaptive capacity of a species, despite significant contractions in its distribution area during the Tertiary period due to prior climate changes. During the Tertiary, L. orientalis persisted in refugia within Anatolia, which provided suitable shelter during periods of climate–induced range reduction. Close to these historical patterns, the model indicates that the suitable shelter areas in Anatolia will likely remain relatively unchanged, especially in valleys on the low shelves of the southern slopes of the western Taurus Mountains. This persistence of suitable habitats indicates an increased likelihood of finding areas conducive to the growth of L. orientalis, even under changing climatic conditions. While the species may hold its ground in certain regions, ongoing monitoring and conservation efforts will be vital to address the declining high suitability areas and the overall suitability landscape for sweetgum trees as climate conditions evolve. Species suitability change The analysis of species suitability change examined both the spatial and proportional gains and losses of species in stable areas, considering both past and future scenarios (Table 4 and Fig. 9). The results showed that the suitable distribution areas of the species varied according to changes in climatic conditions. According to the ca method, the loss rates in the LGM to MH period are 2.5%, absent rates 96.5%, stable areas are 0.8%, gain areas are 0.2%. Under the MH to current period loss areas are 0.2%, absent areas are 96.8%, stable areas are 0.8%, gain areas are 2.2%. Under the current to future period (RCP 2.6/2070), absent areas are 96.8%, stable areas are 1.5%, gain areas are 2.2%. In accordance with the current to future period (RCP 8.5/2070), the loss rate is 2.5%, the area absent is 96.9%, the area stable is 0.5%, and the area gain is 0.1%. According to the wm method, the loss rates in the LGM to MH period are 2.1%, absent rates 96.9%, stable areas are 0.7%, gain areas are 0.2%. Under the MH to current period loss areas are 0.7%, absent areas are 96.8%, stable areas are 0.2%, gain areas are 2.2%. Under the current to future period (RCP 2.6/2070), loss areas are 1.8%, absent areas are 96.9%, stable areas are 1.2%, gain areas are 0.1%. In accordance with the current to future period (RCP 8.5/2070), the loss rate is 2.5%, the area absent is 97.0%, the area stable is 0.5%, and the area gain is 0.1%. Table 4Statistical analysis of the changes in in the suitability of Liquidambar orientalisScenarioLoss %Absent (%)Stable (%)Gain (%)ScenarioLoss %Absent (%)Stable (%)Gain (%)Ensemble TypeCommittee averaging Ensemble TypeWeighted meanLGM–MH2.596.50.80.2LGM–MH2.196.90.70.2MH–Current0.296.80.82.2MH–Current0.796.80.22.2Current–RCP2.6 (2050)1.596.81.50.2Current–RCP2.6 (2050)1.896.91.20.1Current–RCP2.6 (2070)1.497.01.60.0Current–RCP2.6 (2070)1.997.01.10.0Current–RCP8.5 (2050)2.296.90.80.1Current–RCP8.5 (2050)2.297.00.70.0Current–RCP8.5 (2070)2.596.90.50.1Current–RCP8.5 (2070)2.597.00.50.1 Statistical analysis of the changes in in the suitability of Liquidambar orientalis Changes in the predicted suitability of LGM–MH, MH–Current and Current–Future climatic conditions for Liquidambar orientalis based on the committee average and weighted mean It is evident from the analysis of both ensemble methods that suitable habitat areas of the species have decreased from the LGM period to the MH, present and are projected to decrease further from the present to the future. This decline is predicted to manifest in two distinct ways: firstly, a contraction in the spatial extent of suitable growth areas for the species; and secondly, a decline in the category of suitability within these areas (Table 4 and Fig. 9). Discussion Shifts in the Geographic Range of Liquidambar orientalis from the Last Glacial Maximum (LGM) to the Present This study reveals significant insights into the long–term ecological dynamics of Liquidambar orientalis, particularly in relation to environmental changes from the Last Glacial Maximum (LGM) to the present day. Understanding how climatic shifts and habitat conditions influence the distribution of this species enhances our comprehension of plant community responses to changing environments, particularly within the Mediterranean region. The Last Glacial Maximum (LGM) was a period of significant paleoclimatic change, which had a profound impact on the spatiotemporal distribution of global vegetation. This change notably affected thermophilous taxa, such as the Liquidambar genus. Ecological Niche Model (ENM) projections for the LGM indicate that, although the climatically suitable habitat area was quantitatively diminished relative to the present interglacial, it remained spatially extensive. However, it is hypothesised that this modelled potential distribution was subject to significant fragmentation, resulting in substantially reduced realised ecological occupancy. The primary mechanism driving the reduction in ecologically suitable habitat was the combined effect of lower mean temperatures and altered regional precipitation regimes, generating conditions that were suboptimal for this thermophilic angiosperm. This paleoenvironmental transformation caused a significant reduction in the realised geographic range, characterised by the dissolution of the continuous, widespread forest cover typical of warmer periods. Even when evaluated against the seemingly expansive ENM–derived potential range, this interpretation of range contraction is corroborated by, and provides a consistent context for, palynological studies that document the expansion of open–pollen taxa at the expense of forest taxa. For instance, pollen records from Lake Van in eastern Anatolia (Türkiye) spanning the Last Glacial Maximum (LGM) and beyond consistently demonstrate a shift from oak steppe forest during interglacial periods to dwarf shrub steppe and desert steppe during glacial and stadial periods. This confirms the regional retreat of forest cover during colder and drier periods [140, 141]. Similarly, studies of Lake Iznik in northwestern Türkiye show a shift from oak–dominated mesic forest during interglacial periods to steppe during glacial and stadial periods, which supports the interpretation that although forest cover contracted, the landscape likely transitioned toward a more open—but not completely treeless—mosaic, consistent with evidence from western Anatolia indicating low but persistent forest presence during the LGM [142, 143]. These independent records document significant fluctuations and reductions in the diversity and spatial extent of Liquidambar forests in direct response to dynamic LGM paleoenvironmental conditions. Furthermore, it is hypothesised that this contraction facilitated the long–term persistence of the species within climatically buffered, isolated refugia [33]. Post–LGM environmental changes induced a wave of ecological shifts, notably in the Mediterranean region, which is characterized by unique climatic conditions. As a result, the distribution patterns of Liquidambar likely adapted to these changes over time. The initial expansion of L. orientalis in southern Türkiye post–LGM illustrates a response to more favourable climatic conditions that allowed for habitat recovery and species proliferation. This appears to be a common pattern among Mediterranean flora, which often displays resilience and adaptability in the face of environmental shifts. Moreover, regions that exhibited elevated suitability during the LGM, such as the Sandras Mountains and the coastal valleys, should be evaluated within a distinct ecological context in relation to the species’ distributional dynamics. In these areas, temperature does not appear to have been the primary limiting factor. Instead, topographic shielding and the increased precipitation characteristic of the glacial period seem to have played a more decisive role. Glacial geomorphological evidence from the Sandras Mountains indicates that, despite cold conditions during the LGM, the region received higher precipitation than today, with enhanced snowfall contributing to increased regional humidity [144]. Paleoecological records from southwestern and western Türkiye support this interpretation, demonstrating that certain valleys and topographically protected areas maintained effective humidity throughout glacial intervals [143, 145]. These findings are consistent with the substantial contribution of Bio19 in our models and suggest that the micro refugial habitats that ensured the persistence of L. orientalis during the LGM were shaped by the combined influence of precipitation regimes and topographic complexity. Despite early expansion, the temporal decline in suitable habitats for Liquidambar in the Mediterranean region, particularly in southern Türkiye, raises important considerations regarding ecological responses to continued climate change. These findings suggest that while certain regions may experience transient periods of favourable conditions, suitable habitats have fluctuated over time rather than following a unidirectional trend. These fluctuations could be driven by several factors, including increased aridity and altered seasonal climatic patterns that have not supported the sustained growth and spread of Liquidambar forests. These observations echo the broader narrative of climate–induced habitat change, where Mediterranean ecosystems have been subjected to severe alterations over millennia. The historical variability in habitat suitability signifies the ongoing vulnerability of Liquidambar populations, highlighting the importance of maintaining habitat connectivity and resilience amidst shifting climatic conditions. During the Middle Holocene, the Liquidambar orientalis species experienced climatic amelioration (warming and generally wetter conditions) relative to the Last Glacial Maximum (LGM). This led to a period during which it could potentially expand its range or recover from its glacial refugia in southwestern Anatolia [24]. Pollen records indicate a decline in Liquidambar forests during the Holocene compared to the late Pleistocene, suggesting Holocene climatic conditions were less favorable for their development. This pattern of stability followed by decline aligns with increasing human impact. However, a middle Holocene expansion of Liquidambar is confirmed by palynological studies [125], showing an ecological rebound potentially interacting with human pressures. Understanding these historical fluctuations is crucial for interpreting current and future projections of Liquidambar amid ongoing climate change. Our findings emphasize the need for comprehensive conservation strategies. These strategies must consider historical ecological dynamics, adapt to shifting habitats, protect existing populations, restore habitats, and incorporate predictions of future habitat suitability under various climate scenarios. Future research efforts could benefit from a multidisciplinary approach that integrates palaeoecological data with contemporary species distribution models. By conducting more extensive studies on the bioclimatic variables influencing Liquidambar and similar Mediterranean species, researchers could better predict responses to future climate scenarios and their implications for biodiversity. The historical trajectory of L. orientalis in relation to climatic changes offers valuable insights into the adaptability and resilience of Mediterranean flora. Possible geographic regions for L. orientalis in the future Under three different future climate scenarios, the suitability of habitats for Liquidambar will undergo substantial changes compared with the current conditions. The results indicate a significant reduction in potentially suitable habitats for Liquidambar in the Mediterranean region, raising concerns about the survival of this species and the potential risk of extinction attributed to increasing temperatures and greenhouse gas emissions. This points toward a challenging future for Liquidambar in the Mediterranean. In anticipation of a temperature increase of 3 °C to 5 °C in the Mediterranean throughout the twenty-first century, the potential evapotranspiration is projected to average 200 mm annually. This equates to a reduction of 50 mm in annual rainfall [146]. Recent studies, including one concentrating on the Cretan endemic tree Zelkova abelicea, have indicated that the ecological range of numerous so–called 'temperate' species flourishing in Mediterranean mountains might have been underestimated [147]. In southern Türkiye, the anticipated trend suggests a decrease in the potential suitability of habitats for L. orientalis in the future Although minor regional differences appear between scenarios, none of the projections indicate the emergence of new suitable habitats. Instead, all scenarios show a progressive contraction of currently suitable areas, with reductions becoming more pronounced toward 2070. Recent studies focused on the Mediterranean region suggest a likelihood of decreased precipitation, particularly in central and southern areas, with the most significant decline expected in winter. Future climate change will likely have a profound effect on the distribution of Liquidambar in the Mediterranean. Our model outputs show that suitable habitats remain mostly restricted to areas where the species already occurs today, rather than expanding into new regions. This difference may be attributed to the comparatively lower impact of climate change on Liquidambar in southern Türkiye, where the mountains offer localized topographic buffering in certain valleys of the Taurus Mountains. Previous studies on L. orientalis cover the Mid Holocene, present and future (2070) in models. Their model results indicate that the precipitation of the wettest quarter (Bio16) was the most important climatic variable affecting the range shifts for sweetgum trees [24]. In another study modelling the future distribution of the species, precipitation of the wettest month (Bio13) was identified as the variable contributing most strongly to the model [28]. Differences in the bioclimatic variables used, the geographical extent of the study area, and the modelling methodologies applied in previous research can significantly influence the resulting projections. Although both studies focus on the same endemic species, Liquidambar orientalis, they differ fundamentally in terms of research questions, methodological approaches and scientific contributions. Our study analyzes the habitat suitability of this species under two different climate scenarios (RCP 2.6 and RCP 8.5) during five different time periods (Last Glacial Maximum [LGM], Mid Holocene, Present, 2050 and 2070). The modeling process was carried out with an ensemble modeling approach combining 10 different algorithms using the Biomod2 platform and validated with multiple statistical metrics such as ROC, TSS, Kappa and Boyce. In this study, the potential distribution of L. orientalis and the climatic parameters affecting its distribution according to three different scenarios for the Last Glacial Maximum and the Middle Holocene were determined by combining fossil pollen data from the present and the future months of 2050 and 2070. Our results indicate a significant contribution of temperature and precipitation seasonality factors in both models. Therefore, our results suggest that prioritizing mean diurnal range (Bio2) and precipitation of coldest quarter (Bio19) are more influential on the occurrence of Liquidambar orientalis in the Eastern Mediterranean. When the model results obtained in the study were compared, it is possible that the distribution areas of the species in Anatolia were generally lost during the transition from the LGM to the MH, whereas these areas tended to expand during the transition from the MH to the present. However, future projections do not show further expansion. According to the RCP 2.6–2050 and RCP 2.6–2070 scenarios, suitable areas remain mostly stable or undergo slight declines rather than gains. Under the RCP 8.5 scenario, losses increase substantially, indicating a stronger climate–driven contraction. Across both climate scenarios, gains are limited and largely confined to currently occupied regions, consistent with the habitat suitability maps produced (Figs. 5 and 6). Preserving genetic diversity in Liquidambar orientalis is crucial for its future. Genetic studies [133] identified Fethiye Günlüklü, Fethiye İnlice, and Kızılyaka in southwest Türkiye as having the highest genetic diversity (based on allele numbers, genetic diversity, Shannon's constant, and polymorphic loci). These populations should be prioritized in conservation efforts. L. orientalis forests face extinction due to urbanization and agricultural expansion over the past 70 years. Protecting and potentially expanding suitable habitats is vital for the species future distribution. This research, by incorporating climate projections, land use changes, and habitat suitability, aims to provide more accurate predictions for conservation planning. Conclusions The distribution history of Liquidambar orientalis reflects the close interplay between climate dynamics and biogeographic processes in the Eastern Mediterranean. Pollen analysis has revealed that the species has been present in Anatolia since the early Miocene, having survived repeated contractions and expansions of its geographical range in response to climatic fluctuations. The results of the ensemble modelling align with the palaeobotanical record, confirming that Anatolia has long functioned as a key refugium. During the Last Glacial Maximum (LGM), suitable habitats were primarily concentrated in the southwestern coastal belt, while the Mid Holocene period witnessed modest expansions. These findings underscore the remarkable resilience of L. orientalis, which has persisted in its core habitats, including the Köyceğiz basin, despite significant environmental changes. In addition to the primary climatic factors influencing the distribution of Liquidambar orientalis, several other ecological limiting factors must also be considered. The species distribution is constrained by soil characteristics and hydrological regimes, due to its reliance on high groundwater levels, nutrient–rich soils, and permanent water sources [52, 53]. Moreover, the presence of favourable microclimatic conditions in the valleys where the species occurs is critical for maintaining moisture balance. In these areas, topographic factors such as slope and aspect influence both moisture availability and light conditions, thereby affecting local population persistence. Beyond these physical constraints, anthropogenic pressures—such as land use change and resource exploitation—represent significant ecological limitations, contributing to habitat fragmentation and range contraction. Projections of habitat suitability under future climate scenarios suggest contrasting outcomes. The RCP 2.6 scenario indicates that suitable habitats may remain relatively stable, with potential expansion into sheltered mountain valleys and north facing slopes, particularly along the Taurus Mountains. Conversely, RCP 8.5 predicts a substantial decline in areas currently deemed highly suitable, with persistence restricted to climatically stable refugia. These results emphasise that although upslope or northward range shifts may occur, they are limited by both climatic thresholds and the complex, fragmented topography of Anatolia. Consequently, any perceived expansion potential should be interpreted cautiously, taking into account not only climatic projections but also ecological constraints and human induced pressures. Indeed, anthropogenic activities have significantly exacerbated these challenges. Over the past century, L. orientalis forests have undergone severe contraction due to agricultural expansion, urbanisation, and unsustainable harvesting for fuelwood and resin production. Although some suitable habitats may persist in the future, the long–term integrity of these forest ecosystems will depend heavily on the implementation of effective conservation and management strategies. Genetic studies have identified populations in Fethiye (Günlüklü, İnlice, Kızılyaka) as having the highest levels of genetic diversity, designating them as priority areas for protection. The EUFORGEN guidelines and existing practices, including seed stands, clonal seed orchards, and designated nature conservation areas, represent important steps forward. Nevertheless, expanding in situ gene conservation networks—complemented by ex situ efforts—is essential to maintain the species adaptive potential under changing environmental conditions. Several populations of Liquidambar orientalis have been identified within officially protected areas in southwestern Türkiye, including the Köyceğiz–Dalyan and Gökova Special Environmental Protection Areas and Marmaris National Park, in addition to the three EUFGIS registered in situ genetic conservation units (TUR00037, TUR00200, TUR00109 [62–64]. Recent population level genetic analyses indicate low gene flow and a substantial proportion of variance among populations, underscoring the need to maintain multiple, geographically discrete core stands under protection [148, 149]. These protected sites thus serve to safeguard both ecological representation and genetically important units within the species range. In light of its palaeoecological persistence and projected vulnerabilities, L. orientalis should be viewed not only as a relict of past climatic regimes but also as a model species for understanding ecological resilience in Mediterranean forest ecosystems. Conservation efforts targeting this species will therefore contribute to the broader objective of preserving biodiversity and ecosystem functionality in one of the world’s most critical biogeographic hotspots. The long–term survival of the oriental sweetgum ultimately depends on the integration of climate informed projections with conservation planning that acknowledges both ecological and socio–economic realities. By situating L. orientalis within its long temporal legacy and emerging vulnerabilities, this study offers a framework not only for its protection, but also for the broader conservation of Mediterranean relict forests under global change. Acknowledgements We would like to thank the editor and reviewers for their constructive comments and useful recommendations that helped improve the clarity and overall quality of the manuscript. Clinical trial number Not applicable. Authors’ contributions Derya Evrim Koç, Demet Biltekin, Beyza Ustaoğlu, contributed with the collection, organization and analysis of the data and wrote the manuscript draft. Funding No funding. Data availability All the data generated/analyzed during the study (i.e., the environmental variables and occurrence data), are available with the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement All the data generated/analyzed during the study (i.e., the environmental variables and occurrence data), are available with the corresponding author on reasonable request.