Exploring the biodiversity patterns of Anatolia under changing climatic conditions

Abstract

Global change poses an unprecedented threat to Earth's biosphere. Global change induced ecological crises—often described as the sixth mass extinction—have become the most urgent issues for the fields of ecology and conservation science. Particularly, rapidly changing climatic conditions exerts disrupting pressures on global biodiversity. Spatially explicit assessments of biodiversity patterns occupy important roles in the immense challenge facing researchers in understanding and mitigating the effects of climate change on Earth's biosphere. Anatolia, an environmentally complex peninsula characterized by diverse climates and heterogeneous topography, hosts rich biodiversity with high endemism and intersects three biodiversity hotspots. Multiple reports highlight the underprotected status of Anatolian biodiversity and habitats. Furthermore, research on biodiversity patterns in Anatolia remains limited, particularly in terms of spatial assessments, with little to no comprehensive studies available. This study primarily aimed to assess the biodiversity patterns of Anatolia, in the context of climate change. To achieve these goals, species distribution modeling—a widely applied tool in ecology—was employed. Species distribution modeling is a group of mathematical frameworks that relate species occurrences to environmental conditions, thereby allowing researchers to quantify and project species' niches under certain assumptions. These models require predictor variables that are relevant to species' environmental niches. This study focuses on climate change, therefore the 19 bioclimatic variables were utilized as environmental predictors. Bioclimatic variables are various derivations of precipitation and temperature that carry higher biological meaning. The bioclimatic data were sourced from WorldClim 2, an open-access climatic data repository. In order to project biodiversity patterns under climate change, future climate data were obtained from the sixth phase of Coupled Model Intercomparison Project. The averages of the outputs of six different climate models were used to project species distributions for the time periods 2041-2060 and 2081-2100 under the shared socio-economic pathways, SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5. Biodiversity data were sourced from the Global Biodiversity Information Facility, an open-access repository that integrates datasets from various sources, including field studies, museum specimens, and citizen science observations. The dataset used in the study consisted of presence-only species occurrence records for Türkiye, covering amphibians, birds, insects, mammals, and plants. A large portion of these records originated from citizen science observations. Following an extensive data filtering and preparation phase, species distributions were modeled one by one and grouped by higher taxa. To model species distributions, various machine learning and statistical methods were employed. The approaches applied consisted of Maxent, random forests, and boosted regression trees which have been placed as the top performing methods in multimodel benchmarks. Maxent is a Bayesian inference method that utilizes information theoretic concepts to provide the least biased probability distribution of the outcome variable given specified constraints set by the modeler. Random forests are decision-tree based methods, in which multiple trees, that are fitted to bootstrapped subsamples of training data with random subsamples of predictor variables, are combined into an ensemble model. Similarly, boosted regression trees are also decision-tree based models. However, in this approach, simpler trees are iteratively fitted, with each subsequent tree focusing on the remaining unexplained variance. Model fitting process involved a predictor selection phase tailored to each species, split sample cross-validation phase, and a parameter tuning phase before moving on to the fitting of the models. The models were validated on the test sets by their true skill statistics and areas under the receiver operating characteristic curves. Following the calculations of the uncertainty measures, the models were combined into ensembles. Their outputs and projections were utilized to produce biodiversity metrics indicating relative biodiversity and change in biodiversity under different climate scenarios. The spatial assessments of the biodiversity patterns were based on these metrics. Modeling the distributions of large numbers of species from varying taxa opened possibilities for complementary analyses. These included cohabitation analyses in which niche overlap metrics were calculated for pairs of species from different levels of ecological interactions and descriptive analyses of the bioclimatic conditions of areas with varying levels of biodiversity. The results included the distributions of 519 species. By taxa, these species comprised of 5 amphibians, 118 birds, 126 insects, 5 mammalians, and 265 plants. Each taxa displayed their own idiosyncratic biodiversity patterns at finer scales across the study area. Though, the broad patterns shared among the taxa placed the coastal regions of Türkiye as high biodiversity areas. These included the Mediterranean, Aegean, Marmara, and Black Sea regions. At finer scales, Çukurova Delta emerged as one of the areas with the highest levels of biodiversity. The inner Anatolia and southeastern Anatolia were among the regions that exhibited relatively low biodiversity. When the study results were compared to the biodiversity range maps from the International Union for Conservation of Nature (IUCN)—despite fundamental differences in their objectives and methodologies—the two approaches showed considerable agreement. Although each taxon exhibited its own variations, insects showed the greatest deviations from the broad biodiversity patterns shared among taxa. Insect biodiversity was distributed more uniformly across the study area compared to other taxa, with relatively high biodiversity in northeastern Anatolia, where other taxa did not exhibit similar patterns. From a geographic perspective, the findings demonstrated strong associations between high biodiversity patterns and environmental features such as water bodies, forested areas, and topographical complexity. Avian diversity, in particular, exhibited a strong affinity for water bodies. While plants also shared this association, theirs was to a lesser extent. From the bioclimatic perspective, high biodiversity areas showed associations with lower temperature seasonality, higher precipitation, and higher precipitation seasonality. The projections under future climatic conditions indicated general northward and northeastward shifts in biodiversity patterns, with notable inter-taxon variations at finer scales. These shifts intensified and expanded over successive time periods and climate scenarios with increasing radiative forcing. Such findings align with global expectations for biodiversity shifts based on past ecological research. The southern and southwestern coastal regions of Türkiye exhibited declining biodiversity across climate scenarios. The Çukurova Delta, in particular, stood out, showing significant biodiversity losses that became more pronounced under successive climate scenarios. Conversely, the Black Sea region consistently exhibited increasing biodiversity across all taxa, with these patterns intensifying and expanding under successive scenarios. Moreover, complex biodiversity patterns emerged around elevation gradients, water bodies, and forest cover, highlighting the critical roles these features play in shaping biodiversity patterns under changing climatic conditions. These patterns varied considerably depending on taxon, geographic feature, location, and climate scenario. Elevational biodiversity shifts toward higher altitudes were common, congruent with past research. These patterns were often not simple, but rather nuanced. For example, the projections for plant biodiversity showed increases across the Taurus Mountains, except at the highest peaks, where biodiversity declined. Among the uncertainty measures, multivariate environmental similarity surfaces analysis identified southeastern Anatolia as the region that will host novel environmental conditions in the future. The degree of model agreements among themselves varied across taxa. However, broad patterns indicated relatively lower model agreement along the coastal regions, particularly in the southern and southwestern coasts. This study employed species distribution models to map the spatial biodiversity patterns of Türkiye, addressing a notable gap in spatial approaches to biodiversity analyses in Anatolia. The results highlighted the coastal regions of Türkiye as high biodiversity areas and the inner Anatolia and southeastern Anatolia as low biodiversity areas across different taxa. The projections under future climate scenarios indicated biodiversity shifts toward the north and northeast. Additionally, geographical features such as elevation gradients, water bodies, and forest cover were identified as playing critical roles in shaping biodiversity patterns, particularly under changing climatic conditions. Species distribution models have proven to be indispensable tools in ecological research. However, they are not without limitations. Species' responses to environmental change are often complex and multifaceted, yet species distribution models capture only a single aspect of this complexity. This study underscored the need for multidisciplinary approaches that integrate multiple aspects of the issue for a deeper understanding of biodiversity patterns under climate change. In conclusion, this thesis provided a comprehensive spatial assessment of biodiversity patterns in Anatolia in the context of climate change, contributing to our understanding of the biodiversity in Anatolia. This study is expected to support future research, enhance biodiversity assessments, and inform conservation decision-making processes in Türkiye.