High-Resolution Projections of Bioclimatic Variables in Türkiye: Emerging Patterns and Temporal Shifts

Abstract

This study presents a comprehensive spatiotemporal assessment of climatic and bioclimatic conditions across Türkiye for both a historical reference period (1995–2014) and future projections (2020–2099) under two Shared Socioeconomic Pathways (SSP2-4.5 and SSP3-7.0) scenarios using the regional climate model (RCM) COSMO-CLM to downscale large-scale signals to a regional scale at high resolution (0.11). A comparison of the model with ERA5-Land reanalysis data revealed annual biases of +1.41 °C (warm) and −0.28 mm/day (dry), emphasizing the importance of bias correction in regional climate assessments. Bias-corrected future projections indicate a marked warming trend and significant decline in precipitation, especially after the 2060s, with pronounced spatial variability across regions. The most intense warming period of the century is the 2060–2079 period, with an anticipated increase of 0.109 °C/year under the SSP3-7.0 scenario, while, under the SSP2-4.5, it is the 2040–2059 period with an increase of 0.068 °C/year. Bioclimatic variables further illustrate shifts in temperature extremes, seasonal variability, and precipitation patterns. Coastal regions are expected to experience a delay in the onset of wet seasons of 1–2 months, while high-altitude zones show earlier shifts of up to 4 months. Four distinct clusters were identified by using k-means clustering method, each with unique temporal and spatial evolution under both SSP scenarios. Clusters 1 and 2, which predominantly represent continental and interior regions, exhibit a strong association with earlier precipitation onset. Notably, arid and semi-arid conditions expand northward, replacing temperate zones in Central Anatolia. Overall, findings suggest that Türkiye is undergoing a substantial climatic transition toward hotter and drier conditions, regardless of the emission scenario. This study has critical implications for ecological resilience, agricultural sustainability, and water resource management, and offers valuable information for targeted climate adaptation strategies and land-use planning in vulnerable regions of Türkiye.

1. Introduction

Bioclimatic variables are powerful indicators for assessing the impacts of climate change on species. Numerous studies have analyzed these variables at both global [1,2] and regional [3,4] scales using future climate projections. An analysis of future climate scenarios reveals markedly different projections for the mid-century and end-of-century periods; while temperature and precipitation patterns are anticipated to remain relatively stable over various climate models by the mid-century, significant differences emerge towards the end of the century [5], reflecting how the progression of greenhouse gas emissions and policy responses could intensify divergent climatic impacts over time. Hence, the choice of the study period and selected scenario, rather than the selected Earth System Model itself, is the main source of variation, with the size of the shifts in climate conditions increasing over time and under more severe climate change scenarios [6]. Such temporal discrepancies in projections and conditions can result in dramatic effects on bioclimatic variables, ultimately leading to significant changes in species distributions. Studies conducted in Europe predicted that many plant species will face severe threats by the end of century [7].

Intra-annual changes in regional climatic patterns are manifested as temporal shifts in key bioclimatic variables, and the rate and magnitude of these shifts plays a critical role in determining species’ capacities to respond and adapt. Species generally exhibit two primary strategies in response to climate change: migrating to areas with more hospitable habitats to which they are accustomed, or adapting physiologically and phenologically to the new altered environmental conditions [8,9]. The likelihood of successful adaptation is generally higher in regions where climatic changes occur gradually and take place over longer time periods, supporting incremental biological adjustments [9]. However, when the rate of such shifts may surpass the adaptive capacity of species or their dispersal potential, it may result in demographic declines and local extinctions. Moreover, the nature of the emerging conditions resulting from these shifts and whether they will be suitable for adaptation or not remains uncertain.

The trajectory of bioclimatic transformation is not solely dependent on rate of shifts, but also on the nature of the emerging environmental conditions. Numerous studies have projected the regional changes of the climate and the disappearance of previously existing climatic regimes under future climate scenarios. While novel climate conditions may facilitate the introduction of new and/or invasive species, disappearing climate conditions could force some species to abandon their habitats. [10]. Species redistributions, in turn, have the potential to disrupt ecosystem structure and functions, alter ecological interactions, and even affect feedback mechanisms, further influencing regional and global climate systems [8]. Therefore, understanding the nature of both present and projected bioclimatic variables is essential for anticipating ecological and climatological responses. Thus, a detailed understanding of both current and projected bioclimatic variables is crucial for understanding ecological responses, evaluating biodiversity resilience, and developing adaptation and conservation strategies under accelerating climate change.

A range of data sources can be utilized to perform comprehensive bioclimatic assessments. Global datasets such as WorldClim and CHELSA offer widely accessible and standardized bioclimatic variables at various spatial resolutions [11,12]. However, in regions characterized by complex topography and pronounced climatic heterogeneity, these datasets may lack the spatio-temporal resolution necessary to accurately capture localized climatic patterns. For instance, Bedia et al. [13] showed that the observed precipitation patterns are inadequately represented by the WorldClim dataset; in fact, bioclimatic variables related to precipitation have a leaning towards topographical features rather than actual climatic patterns. This misrepresentation can introduce significant uncertainties into climate impact models, particularly those relying on bioclimatic predictors, and complicate efforts to accurately project future climate conditions and associated ecological responses.

Not only topography but also spatial resolution is critically important for bioclimatic variables, as identifying hotspots and seasonal shifts in regional patterns, which can significantly affect species distribution, depends heavily on the spatial precision of the input climate data. According to Bede-Fazekas & Somodi [6], the use of finer-resolution data can significantly change trends that have been seen, especially in mountainous areas where micro- and meso-climatic changes play a significant role. This emphasizes the need for high-resolution climatic data in order to precisely evaluate the effects of climate change on biodiversity.

Also, the topographical features of the regions studied have a great influence on how the bioclimatic variables change, for example, mainland regions associated with higher altitudes are expected to experience greater changes and higher uncertainty in the distant future compared to maritime regions [7,14]. Therefore, an accurate model representation of complex topography, heterogeneous land cover, and irregular coastlines is particularly important for countries with these features, such as Türkiye [15,16]. The enhanced resolution of regional climate models provides more realistic simulations of land–sea interactions, elevation gradients, and small-scale terrain features compared to global-scale models [17,18,19].

Considering all these aspects and limitations, region-specific downscaling techniques and the integration of high-resolution observational datasets are essential, particularly in mountainous or coastal regions where microclimatic variability plays a critical ecological role. Hence, conducting detailed analyses with high spatial resolution model outputs is crucial for effectively understanding how bioclimatic variables influence species distributions in both present and future contexts. Such studies are also essential for identifying climatologically vulnerable areas, which is fundamental for developing regional strategies for climate change adaptation and mitigation [20].

The objective of this study is to conduct a comprehensive spatiotemporal assessment of climatic conditions and bioclimatic variables over Türkiye for both a historical base line period (1995–2014) and future projections extending to the end of the 21st century (2020–2099). This analysis is performed under two Shared Socioeconomic Pathways: SSP2-4.5, representing a moderate mitigation scenario, and SSP3-7.0, reflecting a high-emission, regional rivalry trajectory. High-resolution regional climate model simulations generated by the COSMO-CLM model at 0.11° (∼12 km) spatial resolution are employed to ensure robust representation of local climatic features and topographic influences over Türkiye. By integrating high-resolution climate projections with derived bioclimatic indices, this study aims to characterize both the spatial and temporal variations of climate-driven ecological conditions. This approach facilitates a detailed understanding of how regional climate dynamics may alter habitat suitability, ecological resilience, and species distributions under different future climate pathways, and thereby might provide a scientific basis for climate adaptation planning, biodiversity conservation, and sustainable land-use management strategies in Türkiye. The remainder of this paper is organized as follows: Section 2 describes the model definition with setup options and the study domain. Furthermore, the section elucidates the methodology employed in this study. In Section 3, the results are presented, including the performance of the model, future trends and projections under SSP scenarios, projected shifts of specific periods, and climate sub-regions. Consequently, Section 4 synthesizes the primary conclusions and pertinent discussions.

2. Data and Methodology

2.1. Regional Climate Modelling and Experiment Setup

The assessment of regional scale climate impacts using Earth System Models (ESMs) poses significant challenges due to their relatively coarse spatial resolution, which typically ranges from 100 to 400 km. They are insufficient for accurately representing regional climate heterogeneity and fine-scale processes. Given the mismatch between the resolution of ESMs and the spatial scale required for regional analyses, downscaling becomes indispensable. This limitation necessitates for the utilization of regional climate models (RCMs) to downscale coarse-resolution climate information from ESMs to high-spatial resolutions, appropriate for impact assesments. Therefore, in this study, the climate simulations were conducted via COSMO-CLM [21,22] at a high horizontal grid spacing of 0.11° (∼12 km) over Türkiye (Figure 1) by using initial and boundary conditions from the earth system model developed by a European consortium, EC-Earth3-Veg [23]. The EC-Earth3-Veg was selected for its ability to effectively represent the climate averages of Türkiye, as evidenced by a subsequent analysis that included a comparison of the consistency of various CMIP6 models over Türkiye [24]. In addition to its success in Türkiye, numerous studies examining CMIP6 models for the European domain have demonstrated the superior performance of the EC-Earth3-Veg model [25,26], which includes model components for various physical domains describing the atmosphere, ocean, sea ice, land surface, and atmospheric chemistry, and differs from the EC-Earth3 model [23] through its 2nd generation dynamic global vegetation model called LPJ-GUESS [27] dynamic vegetation. Çetin et al. [28] ascertained that the EC-Earth3-Veg’s June wind spatial pattern, characterized by dynamic vegetation development, exhibits a stronger correlation with both the ERA5 (ECMWF ReAnalysis, 5th generation) and MERRA2 (Modern-Era Retrospective Analysis for Research and Applications, Version 2) reanalysis datasets in terms of their data relating to Türkiye and its vicinity [29].

Türkiye possesses the most extensive surface area in the Eastern Mediterranean region, approximately twice that of Iraq and more than three times that of Romania [30]. In terms of topography, the Marmara region (Northwestern), Aegean coast (Western), and the Southeast Anatolia have relatively low elevations, while elevation increases in the Mediterranean (Southern) and Black Sea (Northern) regions. The Central Anatolia region, on the other hand, is generally flat but characterized by a high-elevation terrain. The eastern part of the Black Sea region and East Anatolia has the highest elevation with complex topography (Figure 1).

Türkiye occupies a unique geographical location, positioned between the temperate and subtropical zones. Coastal regions of the country exhibit relatively milder climate characteristics, owing to the effect of the seas. However, the topographical features of the Northern Anatolian Mountains and the Taurus Mountain Range act as a barrier, preventing inland penetration of maritime effects. Therefore, continental climate characteristics are evident within the interior of the country [31]. The complexity of Türkiye’s geographical features, the spatial orientation of the mountains, and the presence of surrounding seas collectively have led to the emergence of multiple distinct climate regimes, making the study area particularly intriguing for climatic analyses.

To capture this complex structure more realistically, the regional climate model simulation was conducted using the COSMO-CLM by reducing the horizontal resolution to approximately 12 km (∼0.11°). The physical parametrizations for COSMO-CLM were selected based on previous studies conducted for Türkiye at 0.11° spatial resolution [32,33]. In the present study, the model has been updated to the last version (clm6.0), and GLOBCOVER [34] has been utilized as land cover data. Additional details regarding configurations are given in

Table 1. COSMO-CLM model configuration.

Parameter	Value
Resolution	0.11°
Grid numbers	i = 198,j = 128
Vertical levels	40 levels
Dataset	EC-Earth3-Veg [23]
Microphysics scheme	2-Category Ice Scheme
Convection scheme	Tiedke [35]
Radiation scheme	Ritter and Geleyn [36]
Soil and vegetation scheme	TERRA_ML [37]
Land use data	GLOBCOVER [34]
Soil data	FAO-DSMW [38]
Periods	RF: 1995–2014;
	SSP2-4.5 & SSP3-7.0: 2020–2099 1

¹ RF = reference period; SSP2-4.5 & SSP3-7.0 = scenario periods.

.

In this study, SSP2-4.5 and SSP3-7.0 scenarios from Shared Socioeconomic Pathways (SSPs) proposed in the IPCC Sixth Assessment Report [39] were selected to establish alternative future climate pathways for the period of 2020–2099. They are used in climate projections to simulate a range of plausible future trajectories of greenhouse gas emissions and socioeconomic developments. For these purposes, four of the storylines (SSP1, SSP3, SSP4, and SSP5) depict the various combinations of narratives with high or low barriers to adaptation and mitigation, and they were all judged plausible enough to support the creation of SSPs. In this study, the SSP2-4.5 and SSP3-7.0 scenarios were selected from among the SSP scenarios as the focus of this study is to evaluate possible “medium” and “high” emission scenarios for the region and to cover a wide range of future climate projections. The story of SSP2-4.5 describes an intermediate pathway with moderate socioeconomic and environmental trends. On the other hand, the SSP3-7.0 scenario is characterized by regional rivalry and limited international cooperation to reflect a fragmented world where development goals, environmental protection, and human well-being face significant challenges [40].

2.2. Bias Correction

Climate models often simplify real-world conditions, reproducing climate with systematic deviations from the observations. Eventually, they may not accurately depict local climate statistics due to the limitations in spatial resolution or incomplete representation of key physical and dynamical processes. If these systematic biases in simulated climate variables remain uncorrected, they can compromise the credibility and accuracy of following impact analyses, potentially producing unreliable climate risk assessments. Therefore, it is important to compare model outputs with observations and apply an appropriate bias correction method, especially for impact studies.

In this study, we selected the Quantile Delta Mapping (QDM) technique as the bias correction methodology. QDM adjusts systematic biases in quantiles of a modeled series with respect to reference values while maintaining model-projected relative changes in quantiles. Compared to other bias correction methods, QDM has been shown to better preserve relative differences than standard quantile mapping and detrended quantile mapping [41]. In this manner, we employed the QDM method utilizing ERA5-Land reanalysis data with 0.1° spatial resolution as the observational reference. Considering that our model outputs also have a comparable resolution of 0.11°, this choice ensures spatial consistency and improves the accuracy of the bias correction. Nevertheless, the approach has several limitations. The effectiveness of QDM is strongly dependent on the quality of the reference dataset, meaning that uncertainties or structural errors in ERA5-Land are transferred into the corrected outputs. It also assumes that the historical bias structure between model and reference remains stationary under future climate conditions, which may not hold if climate change drives processes outside the historical range [41]. Although QDM improves quantile-based corrections, it can underperform for rare extremes due to limited sampling and may produce artificial results when extrapolating beyond observed values. Furthermore, corrections are applied independently at each grid cell and time step, which may disrupt spatial coherence and temporal persistence, and the method typically treats variables univariately, potentially neglecting physical interdependencies such as temperature–precipitation coupling. Finally, while QDM adjusts the statistical distribution of climate variables, it does not address underlying physical errors in the regional climate models, meaning that the corrected data may appear more realistic but still lack full dynamical consistency [42].

In our case, QDM greatly improves the model outputs; therefore, the confidence over the future projections is improved. While bias analysis showcases the warm and dry tendencies of COSMO over Türkiye during the reference period, the QDM results clearly show how these tendencies are alleviated. For temperature, the model exhibits a clear warm bias in every season (area-wide mean +2.27 °C vs. ERA5; median +2.03 °C), which is exactly what the CDFs show as a right-shifted curve and the PDFs as peaks at higher values with a slightly narrower spread. QDM substantially recenters the distribution, cutting the mean bias to +0.86 °C and the median to ≈+0.73 °C, and the CDF/PDF curves nearly overlap ERA5 through the interquartile range (IQR). Improvements extend to both tails: the cold tail that is too warm in the raw model (p5 ≈ +1.35 °C) is brought close to neutral with QDM (average p5 −0.07 °C), while a small residual warm skew persists aloft (QDM p95 +1.80 °C), most visible in DJF. Seasonally, QDM removes the warm bias most in MAM and JJA (mean +0.76/+0.63 °C), with a larger winter residual (DJF mean +1.14 °C). The box-plots mirror this narrative, as model medians are above ERA5 with slightly tighter IQRs, whereas QDM pulls medians toward ERA5 and restores the spread, leaving only a modest high-end offset in winter. Spatial variability remains close to ERA5 in both raw and QDM (std 0.56 vs. 0.53 °C), indicating broadly similar spatial gradients after correction (given in the Supplementary Materials as Figures S1–S3).

For precipitation, the raw model is generally too dry (mean −0.47 mm/day overall), under-dispersed (std −0.27 mm/day relative to ERA5), and dominated by low precipitation amounts, as indicated by CDFs that rise early and PDFs that are concentrated at low intensities. QDM largely removes the dryness (mean −0.06 mm/day), recenters the median, and restores variability (std +0.35 mm/day vs. ERA5). The high tail reveals a seasonal contrast: QDM strengthens high precipitation amounts in DJF/MAM (p95 above ERA5), but remains somewhat thin in the warm season, with p95 below ERA5 in JJA/SON (largest shortfall in JJA). Box-plots capture the same shift. Model medians and IQRs lie below ERA5 with short upper whiskers, while QDM raises medians and lengthens upper whiskers, bringing the distribution much closer to ERA5 overall. Netting across seasons, QDM’s p95 ends slightly above ERA5 on average, as wintertime corrections outweigh summer deficits, marking a substantial recovery of distributional shape and magnitude compared to the raw model outputs (given in Supplementary as Figures S4–S6).

2.3. Bioclimatic Variables

Bioclimatic predictors, which are derived from long term climate data, indicate biologically relevant aspect of temperature and precipitation patterns. They are widely used in ecological modeling, agricultural applications, biodiversity research, and land-use planning to analyze and understand the impact of climate change. They provide information on seasonal mean climate conditions, intra-year seasonality (temperature of the coldest and warmest months, precipitation of the wettest and driest quarters), and annual conditions (annual mean temperature, annual precipitation, annual range in temperature and precipitation). To understand the past, present, and future distributions of species, we can benefit from the strong patterns provided by bioclimatic variables [43].

We calculated the bioclimatic variables for the present and future periods using high-resolution COSMO model simulations of maximum, minimum, and 2 m temperatures, and total precipitation data, following the formulations described in O’Donnell and Ignizio [43], and implemented the calculations through NCL scripts. A list of all 19 bioclimatic variables and their units is given in

Table 2. Bioclimatic variables and their units.

Code	Name	Unit
BIO1	Annual Mean Temperature	°C
BIO2	Mean Diurnal Range	°C
BIO3	Isothermality	%
BIO4	Temperature Seasonality	%
BIO5	Max Temperature of Warmest Month	°C
BIO6	Min Temperature of Coldest Month	°C
BIO7	Temperature Annual Range	°C
BIO8	Mean Temperature of Wettest Quarter	°C
BIO9	Mean Temperature of Driest Quarter	°C
BIO10	Mean Temperature of Warmest Quarter	°C
BIO11	Mean Temperature of Coldest Quarter	°C
BIO12	Annual Precipitation	mm
BIO13	Precipitation of Wettest Month	mm
BIO14	Precipitation of Driest Month	mm
BIO15	Precipitation Seasonality	%
BIO16	Precipitation of Wettest Quarter	mm
BIO17	Precipitation of Driest Quarter	mm
BIO18	Precipitation of Warmest Quarter	mm
BIO19	Precipitation of Coldest Quarter	mm

.

3. Results

3.1. Evaluation of COSMO-CLM Performance and Bias Correction

We evaluated the EC-Earth3-Veg-driven COSMO-CLM simulations by comparing them with ERA5-Land gridded reanalysis data to assess the performance of the model in capturing climate characteristics of Türkiye. We focus mainly on mean 2 m temperature and total precipitation (Figure 2). To ensure spatial consistency, COSMO-CLM simulated data were bilinearly interpolated to align with the ERA5-Land grid resolution.

The comparative analysis for the reference period (Figure 2) reveals spatially coherent patterns between the two datasets, revealing that the model effectively captures the spatial temperature gradient shaped by Türkiye’s complex topography and surrounded seas. According to the COSMO-CLM simulations, the highest temperatures are found in the Aegean and Southeastern Anatolia, while the lowest values occur in the Black Sea and Eastern Anatolia. The model estimates a national mean annual temperature of 12.24 °C, with extremes ranging from 2.04 °C to 21.38 °C over Türkiye. In comparison, ERA5-Land reanalysis data indicate slightly lower values, with a mean of 11.04 °C, a minimum of 0.04 °C, and a maximum of 19.79 °C.

On average, COSMO-CLM tends to overestimate temperature across most of Türkiye, and this is particularly pronounced in the interior and eastern parts of the country at high elevations (Figure 2, upper panels). These systematic biases are further supported by the quantitative analysis summarized in

Table 3. Annual and seasonal values for COSMO-CLM, ERA5-Land, and their differences for the reference (1995–2014) period.

	Temperature [°C]			Precipitation [mm/day]
Period	COSMO-CLM	ERA5-Land	COSMO-ERA5	COSMO-CLM	ERA5-Land	COSMO-ERA5
Ann	12.24	11.04	1.19	0.49	0.64	−0.16
DJF	1.07	0.59	0.48	0.87	0.88	0.00
MAM	10.90	9.48	1.42	0.61	0.81	−0.20
JJA	23.25	21.73	1.52	0.11	0.30	−0.19
SON	13.72	12.38	1.35	0.35	0.59	−0.24

, which presents detailed annual and seasonal temperature and precipitation statistics for both COSMO-CLM and ERA5-Land, including the four climatological seasons: DJF (winter), MAM (spring), JJA (summer), and SON (autumn). The mean annual temperature difference between the two datasets is found to be 1.19 °C, with the warmest bias observed during spring (JJA, +1.52 °C) and the lowest in winter (DJF, +0.48 °C). The local discrepancies in the annual means varying between −1.45 °C and +4.93 °C. These deviations are particularly notable in regions characterized by complex topography, where the model’s spatial resolution and terrain representation can compromise model performance. These results underscore the importance of regional validation, particularly in topographically diverse areas where model–observation differences are more pronounced.

The bottom panels of Figure 2 compare the mean annual precipitation simulated by the COSMO-CLM model with ERA5-Land reanalysis data for the reference period. Although the COSMO-CLM model performs reasonably well in capturing Türkiye’s broad precipitation patterns with maximum precipitation along the eastern Black Sea coast and lower values across central and southeastern Türkiye, it generally underestimates total precipitation, while modest overestimations occur especially in regions with complex terrain. This underestimation is most pronounced along the Black Sea coast, where the model simulates substantially drier conditions (less than −1 mm/day) compared to observations. On the contrary, the model overestimates the precipitation over the Antalya region when compared to the reanalysis data. The tabulated results confirm these spatial tendencies, with COSMO-CLM underestimating annual mean precipitation by approximately −0.16 mm/day with almost no seasonal tendencies. These modest yet consistent biases suggest that, although COSMO-CLM performs reasonably well in simulating the general precipitation climatology, regional and seasonal differences are likely influenced by complex topography and mesoscale circulation features.

In general, the simulation results illustrate that the regional climate model captures the broad spatiotemporal patterns of temperature and precipitation over Türkiye. However, systematic biases, especially the warm temperature bias, require the application of bias correction techniques prior to calculating bioclimatic variables for the reference and future periods under SSP2-4.5 and SSP3-7.0 scenarios. Therefore, the QDM bias correction technique should be applied to the COSMO-CLM outputs of both the reference and future periods to improve the model’s ability to represent the influence of topography and produce more realistic temperature and precipitation variations.

3.2. Projected Temperature and Precipitation Trends Under SSP Scenarios

Following bias correction, a trend analysis was conducted to evaluate long-term changes in climate indicators. Identifying warming, cooling, drying, or wetting trends is essential for understanding both the magnitude and direction of climate change at the regional scale. For this purpose, Sen’s slope estimator [44] was employed as a non-parametric method to quantify monotonic tendencies in temperature and precipitation time series for the reference period and twenty years period in the future for the SSP2-4.5 and SSP3-7.0 scenarios. This method works particularly well for climate data because it is not affected by outliers and does not require the data to follow a normal distribution. In addition, a 95% confidence interval was applied to assess the statistical significance of the estimated trends, allowing for a clear distinction between meaningful climatic shifts and natural variability.

A robust signal of country-wide warming is the most consistent projection across both scenarios, though the rate of this warming unfolds differently over the century. Figure 3 presents the spatial distribution of temperature (left) and precipitation (right) trends for the reference and future periods for the SSP3-7.0 scenario (SSP2-4.5 is given in Supplementary as Figure S7). The SSP3-7.0 scenario shows a relatively steady intensification of warming, with the trend becoming progressively stronger toward the end of the century. In the early period (2020–2039), the Marmara region (northwest) is projected to experience relatively lower temperature increases compared to the national average (+0.015 °C/year). However, warming in this region is expected to accelerate in 2040–2059 and 2060–2079, before slightly accelerating over the Thrace region toward the end of the century. This temporal pattern of intensifying and then moderating warming is consistent across the country, with the strongest overall temperature increase projected for Eastern Anatolia. Countrywide trends for the 2040–2059 period jump to +0.070 °C/year, reach a maximum for the future periods during the 2060–2079 period with an increase 0.109 °C/year, and slow down for the 2080–2099 period to around +0.083 °C/year.

In contrast, the SSP2-4.5 scenario presents a more variable warming pattern. A notable feature is a temporary slowdown of warming during the 2020–2039 period, where the average trend is a modest +0.017 °C/year and is statistically insignificant for most of the country (2.0% of area). However, this is immediately followed by a period of dramatic acceleration (2040–2059), which emerges as the most intense warming period of the century under this scenario. The average warming rate surges to +0.068 °C/year, with the trend becoming statistically significant across the entire nation (100%). After a period of moderation from 2060 to 2079, a second intense warming pulse occurs toward the end of the century (2080–2099), with a high average rate of +0.064 °C/year. Despite these differences in temporal patterns, the spatial distribution of warming is consistent. Both scenarios identify Eastern Anatolia as the region experiencing the most pronounced and rapid temperature increases, reflecting the change in the precipitation type in the future and decreased snow cover areas during the cold season.

The divergent pathways between the scenarios are even more pronounced for total precipitation. Under the SSP3-7.0 scenario (Figure 3, right panel), the primary narrative is a progressive drying that becomes particularly acute after the 2060s. Except for the modest increase in precipitation for the Black Sea region during the 2020–2039 period, the general trend is −0.34 mm/year. The countrywide trend for precipitation for the 2040s is −0.22 mm/year, which is slower than the previous period, most likely due to the localized increase in rainfall in Southeastern Anatolia, which is likely influenced by moisture transport from the Persian Gulf. However, the most pronounced negative precipitation trend is projected for 2060–2079 (−0.82 mm/year), with some moderation anticipated by the end of the century (−0.37 mm/year), except the Adana region, where positive values are apparent.

Under the SSP2-4.5 scenario, precipitation trends exhibit a distinct temporal pattern, with severe and widespread drying projected in the early century, a temporary stabilization mid-century, and a renewed intensification of drying, especially in the south and southeast, toward the century’s end.

These scenario-dependent trends highlight the complexity of Türkiye’s future climate. The persistent warming trend indicates inevitable heat-related stress, while the varying precipitation patterns underscore significant uncertainties regarding water resource availability. The SSP2-4.5 scenario, however, reveals a notably different temporal evolution of precipitation trends compared to higher emission scenario. The SSP2-4.5 scenario indicates an initial, severe drying event, succeeded by an increase in aridity later in the century, while the SSP3-7.0 scenario proposes a more gradual transition towards aridity. This temporal variability underlines the importance of nonlinear climate responses in estimation of future water availability and associated risks.

3.3. Spatial Patterns of Bioclimatic Variables

The habitats and life cycles of plant and animal species are impacted by climate change. Temperature and precipitation trends reveal information about how fast and well these species can adapt. Trends can reveal crucial details about a species’ persistence and adaptability when analyzed in combination with bioclimatic variables. Therefore, we calculated the 19 bioclimatic variables and derived their statistical averages for two future scenarios: SSP2-4.5 and SSP3-7.0 (

Table 4. Bioclimatic variable values (one-decimal) for reference and future periods under SSP2-4.5 and SSP3-7.0.

Variable	REF	2020–2039		2040–2059		2060–2079		2080–2099
		SSP2	SSP3	SSP2	SSP3	SSP2	SSP3	SSP2	SSP3
BIO1	10.7	11.6	11.5	12.2	12.2	12.9	13.7	13.7	15.2
BIO2	9.4	9.7	9.6	9.6	9.6	9.7	9.6	9.7	9.9
BIO3	28.9	29.2	28.5	28.9	28.4	29.1	28.1	29.6	28.7
BIO4	8.6	8.9	8.8	8.9	9.1	8.8	9.0	8.5	9.1
BIO5	28.7	29.9	30.1	30.8	31.4	31.4	32.9	32.0	34.5
BIO6	–3.8	–3.2	–3.6	–2.6	–2.2	–2.0	–1.2	–0.7	0.2
BIO7	32.5	33.0	33.6	33.4	33.6	33.4	34.1	32.7	34.3
BIO8	5.4	6.0	5.9	6.8	5.9	6.5	7.5	8.2	7.7
BIO9	20.6	22.0	21.7	22.7	23.0	23.4	24.5	23.9	26.2
BIO10	21.3	22.3	22.4	23.1	23.4	23.7	24.9	24.2	26.5
BIO11	0.4	0.8	1.1	1.7	1.5	2.4	3.0	3.7	4.5
BIO12	753.4	677.8	711.8	689.0	708.5	647.4	654.8	629.3	539.0
BIO13	140.6	124.4	128.4	133.2	135.8	121.0	122.8	113.3	104.2
BIO14	6.4	5.0	5.9	4.3	4.7	3.2	3.2	3.4	2.7
BIO15	70.8	72.0	70.2	73.4	72.7	74.6	74.4	70.8	77.0
BIO16	338.1	307.0	314.5	309.4	318.8	294.7	302.1	278.3	250.2
BIO17	34.8	24.6	32.1	27.4	26.8	19.6	21.1	21.6	17.3
BIO18	72.5	44.2	62.9	51.2	56.2	36.9	42.6	47.7	36.8
BIO19	263.8	238.6	241.2	218.6	246.1	236.1	239.0	219.6	211.3

).

The projected evolution of bioclimatic variables under both SSP2-4.5 and SSP3-7.0 scenarios illustrates a robust and consistent signal of warming and hydroclimatic alteration across Türkiye. Figure 4 highlights the spatial and temporal distribution of BIO1 (Annual Mean Temperature) under SSP3-7.0, which serves as a central indicator in understanding Türkiye’s climate trajectory based on COSMO-CLM model outputs. During the reference period, the lowest annual mean temperatures were observed over the Black Sea and Eastern Anatolia, while the highest occurred in Southeastern Anatolia and the Aegean region. The model simulations show that annual mean temperatures range from −0.09 °C to 19.64 °C, with an average of 10.69 °C. BIO1 increases steadily under both scenarios, reaching 13.68 °C by 2080–2099 under SSP2-4.5 and a more pronounced increase (15.17 °C) under SSP3-7.0, indicating a stronger warming trajectory in the high-emissions. The temperature rise during each period is more noticeable in the southern latitudes compared to the northern latitudes and over the elevated topographical reagions in eastern Türkiye. This warming is further reflected in the maximum temperature of the warmest month (BIO5), which rises from a reference value of from 28.70 °C to 31.99 °C (SSP2-4.5) and 34.46 °C (SSP3-7.0) by the end of the century, and in the minimum temperature of the coldest month (BIO6), which increases from −3.83 °C to −0.70 °C (SSP2-4.5) and +0.16 °C (SSP3-7.0), indicating a significant reduction in cold extremes, especially under SSP3-7.0.

Supporting this key finding, the BIO2 (Mean Diurnal Range), which corresponds to the average daily temperature range, shows minimal increase under both scenarios with decadal variability. However, these changes are spatially coherent, particularly in inland and high-altitude regions where the warming signal is more pronounced. Overall, BIO2 demonstrates a small change along coastal areas due to thermal effect of the seas in proximity, controlling the day and night temperature variability, and illustrates a notable increase over the continental interiors and mountainous zone as a result of the differential warming trend reflected in BIO1.

On the other hand, the changes in BIO3 (Isothermality), calculated as the ratio of BIO2 to BIO7, further contextualize the evolving temperature dynamics in different time scales. A notable decline in isothermality is observed across western Türkiye and coastal zones, indicating mild temperatures due to a steeper increase in annual temperature range (BIO7) compared to daily temperature fluctuations (BIO2). This is mostly due to the fact of the perpendicular orientation of mountain ranges to the coastline in the Aegean region facilitates the penetration of warm westerly winds into the Central Anatolia, resulting in milder winter conditions. In contrast, Eastern Anatolia exhibits a sharp increase in BIO3 by the 2080–2099 period, driven primarily by a significant late-century rise in daily temperature ranges, amplifying the BIO1 warming signal in the region through enhanced temperature stability. Projections for both scenarios reveal asymmetry in warming rates of the seasons, with increasing BIO5 (Maximum Temperature of Warmest Month) steeper than BIO6 (Minimum Temperature of Coldest Month), highlighting that, summers are warming faster than winters. This discrepancy contributes to higher BIO7 values in the west, which in turn affects BIO3 and underlines the uneven regional response to warming.

BIO4 (Temperature Seasonality), which quantifies the variability of monthly temperatures, also reinforces the BIO1 pattern. While regional distribution remains relatively consistent with the highest seasonality in Eastern and Southeastern Anatolia, the temporal trend reveals a growing seasonal temperature variability in western Türkiye. This trend parallels the rising annual means (BIO1), indicating that warming is accompanied by more variable temperatures over time in these regions. However, the mean temperature of the coldest quarter (BIO11) displays a significant rise, especially under SSP3-7.0, increasing from 0.39 °C to 4.53 °C, which implies substantial winter warming and likely reduction in cold days and snow cover.

While temperature-related bioclimatic variables reveal a consistent and spatially diverse warming pattern across Türkiye, it is equally important to examine how precipitation-related bioclimatic variables respond under similar future scenarios. Annual precipitation (BIO12) declines from 753.4 mm to 629.3 mm under SSP2-4.5 and further to 539.0 mm under SSP3-7.0 by the late century, reflecting a clear reduction in total moisture availability in Türkiye. Figure 5 illustrates the projected changes in BIO12 (Annual Total Precipitation) across Türkiye, based on COSMO-CLM simulations under the SSP3-7.0 scenario. During the early future periods (2020–2039 and 2040–2059), the model suggests a modest increase in annual precipitation, particularly in the northeastern regions. However, starting from the 2060s, a clear drying trend emerges, culminating in a country-wide reduction of up to 220 mm by the end of the century, as also indicated in Table 4. These shifts reveal BIO12’s role as a critical driver of hydroclimatic transformation under future climate scenarios. Supporting this pattern, BIO13 (Precipitation of Wettest Month) maintains a consistent spatial structure, with maximum values along the eastern Black Sea coast and minima in Central Anatolia throughout all periods. However, while early-century projections indicate regional increase along narrow coastal stretches of the Black Sea, a widespread decrease in wettest-month precipitation is anticipated toward the end of the century. The average BIO13 value declines significantly under the SSP3-7.0 scenario, from 140 mm to 104 mm (113 mm for SSP2-4.5), reflecting the overall drying signaled by BIO12. A similar declining pattern exists for both scenarios for precipitation of driest month (BIO14). Initially, a slight increase is projected in parts of the western and central Black Sea region, but, by the mid-century and beyond, this turns into a consistent decline, especially across eastern Marmara and the broader Black Sea basin.

Complementing these findings, BIO15 (Precipitation Seasonality) provides comprehension of intra-annual variability. High BIO15 values reflect more erratic precipitation distributions throughout the year. BIO15 increases gradually but steadily in both scenarios, with a more pronounced rise under SSP3-7.0 (up to 76.96%) compared to SSP2-4.5 (up to 74.64%), indicating greater variability in monthly precipitation patterns. The simulations suggest increasing precipitation seasonality in Marmara, Central Anatolia, and Southeastern Anatolia, likely due to sharper contrasts between wet and dry months. In contrast, Antalya shows a decline in seasonality, which may be attributed to the expected reduction in overall precipitation (BIO12) that smooths out the monthly extremes.

Taken together, these bioclimatic variables portray a coherent narrative: Türkiye’s future precipitation regime will become drier and more uneven, with the most dramatic decreases occurring in the latter half of the century. While regional nuances exist, BIO13, BIO14, and BIO15 collectively reinforce the long-term drying signal captured by BIO12, with potentially profound implications for water availability, agriculture, and ecosystem stability. Notably, precipitation amounts during the warmest quarter (BIO18) and the coldest quarter (BIO19) decline significantly under both scenarios, implicating increased drought risk during the growing season. The bioclimatic variables calculated under the SSP2-4.5 scenario exhibited a similar spatial distribution to those under the SSP3-7.0 scenario; however, the statistical values were found to be more moderate. Due to the large number of figures, the bioclimatic variables that could not be included in the main text of the manuscript can be found in the Supplementary Materials as Figures S8–S11.

3.4. Projected Shifts of Specific Periods

When analyzing bioclimatic variables, quarterly indices are often overlooked due to their grid-specific calculation method, which can complicate interpretation. Nonetheless, their importance becomes evident when considering the temporal shifts in precipitation seasonality, which can shift by from one to six month and reveal meaningful changes in seasonal climate behavior [6,45].

To investigate these shifts in maximum total precipitation, we calculated cumulative precipitation over a moving three-month window for each study period and identified the wettest quarter at each grid point under the SSP2-4.5 (given in Supplementary Figure S12) and SSP3-7.0 (Figure 6) scenarios. The left panel of Figure 6 illustrates the onset month of the wettest quarter, coded numerically from 1 (January) to 12 (December). The reference period (1995–2014) show a clear geographic distinction between low-altitude coastal regions along the Marmara and Aegean seas, which typically experience a winter-dominant precipitation regime starting in November or December, and the high-altitude continental regions, especially eastern Anatolia, which has a spring-dominant regime with precipitation onset in March–April.

Future projections from both scenarios indicate a consistent disruption of this seasonal pattern. Coastal zones, especially along the Aegean Sea, are projected to experience a delay in their seasonal precipitation peaks by from one to two months (a positive shift). Conversely, the vast continental and mountainous regions are expected to see a significant advancement in the timing of maximum precipitation period, which suggests a systematic earlier onset of their wettest season (a negative shift). The right panel of Figure 6 displays the difference in the onset of the wettest quarter compared to the reference for SSP3-7.0 scenario. Negative values (shades of blue) signify a shift toward earlier months, while positive values (shades of red) indicate a shift toward later months. Along coastal zones, positive shifts dominate, suggesting a delay in seasonal precipitation peaks. Conversely, in the mountainous regions, negative shifts can last up to four months in some areas. This asymmetric pattern indicates that while positive shifts are generally limited to one-month, negative shifts are broader and more substantial, concentrated in elevated terrains.

While this asymmetric pattern is a robust finding, a direct comparison of the difference plots reveals that the magnitude and spatial extent of this reorganization are considerably more pronounced under the SSP2-4.5 scenario. The advance of the wet season in the Anatolian plateau becomes more intense and extends up to four months earlier in many areas by the end of the century. Furthermore, the spatial expansion of this earlier, winter-dominant regime is more aggressive under SSP2-4.5, pushing progressively westward and covering a larger contiguous area of the Anatolian plateau compared to SSP3-7.0.

Among the bioclimatic variables, the mean temperature of the wettest quarter (BIO8) shows a notable similarity with the shifts in the onset month of the wettest three-month period (Figure 7). The consistency between changes in BIO8 and shifts in the onset month of the wettest quarter is likely attributed to the seasonal reorganization of precipitation regimes under climate change. In future periods, regions where BIO8 increases (positive values—reddish) tend to correspond with positive shifts in the timing of the wettest quarter, while regions where BIO8 decreases (negative values—blue) show a pattern similar to negative shifts in timing. In other words, the mean temperatures of the wettest quarter have shifted towards cooler months with negative shifts and towards warmer months with positive shifts.

Beyond changes in annual totals, a fundamental reorganization of precipitation seasonality is projected across Türkiye. An analysis of the onset month of the wettest quarter reveals a remarkably robust, yet asymmetric, pattern of change that is consistent across both the SSP2-4.5 and SSP3-7.0 scenarios. This powerful signal in both scenarios points to a fundamental restructuring of Türkiye’s hydroclimatic system. The accelerated shift in wettest-quarter precipitation timing in high-elevation zones signals a increased climatic sensitivity of these areas to future climate perturbations. The pronounced shift toward an earlier and often drier season in the continental area might have profound implications for water resource management, agricultural cycles, and ecosystem phenology.

3.5. Climate Sub-Regions

We identified the climate sub-regions of Türkiye combined for monthly temperature and precipitation for the reference period and future projection periods for SSP2-4.5 (given in Supplementary Figure S13 and Table S1) and SSP3-7.0 scenarios with 20-year intervals, using the k-means clustering method implemented in Python 3.12.7 via the scikit-learn package [46]. To determine the optimal number of clusters, we applied the elbow and silhouette methods. The elbow method was applied by evaluating the inertia values from scikit-learn’s KMeans outputs, and the silhouette method was computed using silhouette_score from scikit-learn, resulting in four clusters. The cluster maps (Figure 8) show a clear spatial reorganization of Türkiye’s climatic zones across the reference and future periods under the SSP3-7.0 scenario. Each cluster represents a unique climate regime shaped by the joint seasonal variability of temperature and precipitation, and their changes over time emphasize shifts in regional climate dynamics. Characteristics for all the clusters, cluster average values of temperature and precipitation, their standard deviations, and the percentage of the grid area each cluster covers, are given in

Table 5. Mean and standard deviation values and percentage of total area for clusters under SSP3-7.0 scenario.

Cluster	Period	Temperature [C]	T_stddev [C]	Precipitation [mm]	P_stddev [mm]	Area %
1	1995–2014	5.56	1.70	1.69	0.41	21.57
	2020–2039	6.38	1.76	1.95	0.51	22.36
	2040–2059	7.20	1.75	1.92	0.48	22.73
	2060–2079	8.74	1.69	1.78	0.43	21.97
	2080–2099	10.51	1.65	1.45	0.38	23.14
2	1995–2014	10.40	1.47	1.35	0.35	33.31
	2020–2039	11.35	1.31	1.60	0.48	37.39
	2040–2059	12.09	1.29	1.60	0.48	38.51
	2060–2079	13.51	1.25	1.41	0.42	39.14
	2080–2099	15.13	1.23	1.19	0.41	43.50
3	1995–2014	10.83	3.38	2.92	0.62	11.79
	2020–2039	10.01	3.31	3.93	1.13	8.92
	2040–2059	10.49	3.27	3.98	1.16	8.15
	2060–2079	12.47	3.35	3.40	0.97	9.76
	2080–2099	12.56	3.36	3.38	1.00	6.02
4	1995–2014	14.24	2.02	1.05	0.33	33.32
	2020–2039	15.51	1.85	1.39	0.63	31.33
	2040–2059	16.23	1.85	1.41	0.63	30.61
	2060–2079	17.74	1.78	1.38	0.61	29.13
	2080–2099	19.38	1.71	1.14	0.52	27.35

.

Cluster 1 (green) represents a sub-climate type characterized by the lowest average temperatures and moderate precipitation, covering the inland areas of the Black Sea region and Eastern Anatolia. Temperatures in this cluster are projected to rise progressively from 5.56 °C to 10.51 °C by the end of the century, while average precipitation is expected to decline from 1.69 mm to 1.45 mm. The relatively stable standard deviation of temperature suggests consistent seasonal variability. Initially, there was increased precipitation in the mid-century but a decrease afterward, along with decreasing interannual variability, which might be the result of a shift toward a less frequent but more stable precipitation regime. In general, this cluster remains relatively stable and its spatial extent increase from around 21.57% to 23.14% throughout the century.

Cluster 2 (blue) describes a warm and moderate precipitation sub-climate type, predominantly observed mostly in Central Anatolia and continental interiors. The cluster’s temperature continuously increases while its precipitation first increases slightly until the mid-century and is then followed by a steady decline. It currently covers one-third of Türkiye, about 33.31% during the reference period, and it is projected to expand to 43.5% by the end of the century. It replaces more humid and temperate zones which result in a dominance of drying.

Cluster 3 (gray) encompasses moderate temperature regions with highest precipitation, including parts of Black Sea, the Aegean, Mediterranean, and northern Southeastern Anatolia. While it covers the smallest area in the country, approximately 11.79% of the land area in the reference period, its extent is expected to decrease to around 6.02% by the century’s end. Interestingly, outside the Black Sea region, this sub-climate type is expected to diminish/disappear in all future periods. This contraction is probably attributed to warming that increases with elevation, where areas formerly characterized by snow dominance shift into warmer and drier climatic regimes. These findings are consistent with the broader sensitivity of mountain ecosystems to climate change.

Cluster 4 (yellow) corresponds to a warm and dry Mediterranean climate and covers Central Anatolia, southern parts of the Marmara region, the Aegean region, and the southern parts of Southeastern Anatolia. This cluster currently covers around 33.32% of the area, but it is projected to shrink to 27.35% by the end of the century. It shows the largest temperature increase among all clusters. Precipitation declines in the cluster but variability remains high.

Under the SSP3-7.0 scenario, all clusters exhibit strong warming with the least increases occurring in Cluster 3. Precipitation exibits an increasingly heterogeneous pattern in time. While some zones, such as those described by Cluster 3, continue to receive a high level of precipitation in the future, they shrink in terms of area coverage. Cluster 2 becomes dominant, while Clusters 3 and 4 shrink, suggesting homogenization of climate regimes and a potential loss of climatic diversity.

The joint assessment of projected climate regimes and seasonal precipitation timing under both the SSP3-7.0 and SSP2-4.5 scenarios indicates a coherent pattern of climatic transition across Türkiye. Even though the magnitude and timing of the changes vary depending on the scenarios, both projections consistently indicate a substantial northward expansion of arid and semi-arid conditions, primarily replacing temperate climates. Moreover, the spatial resemblance between cluster boundaries and shifts in the onset of the wettest quarter points out distinct regional climate responses. For instance, Clusters 1 and 2, mainly representing continental and interior regions, are strongly associated with earlier precipitation onset (negative shift), while Cluster 3 exhibits relative temporal stability, and Cluster 4 aligns with delayed or unchanged seasonal precipitation timing. These findings emphasize the fact that Türkiye’s principle agro-ecological zones are highly vulnerable to climate change, irrespective of the pathway.

This underscores the robustness of the projection that Türkiye’s agricultural and ecological heartland is highly vulnerable to climate change, regardless of the scenario.

Comparing the sub-climate types identified through K-means clustering (Figure 8) with the seasonal shift in the three months of maximum precipitation (Figure 6), it becomes evident that Clusters 1 and 2 correspond to areas with high negative shift values, while Cluster 3 generally aligns with regions showing near-zero or slightly positive shifts. Cluster 4 is associated with areas experiencing either zero or positive shifts. The SSP2-4.5 scenario shows seasonal shifts generally similar to SSP3. The main exception is the mountain ranges in south-southeast Anatolia, which experience positive shifts under SSP2-4.5, unlike in the SSP3-7.0 scenario.

4. Conclusions and Discussions

This study presents a comprehensive spatiotemporal assessment of climatic and bioclimatic conditions across Türkiye, covering both a historical baseline period (1995–2014) and future projections (2020–2099) under two Shared Socioeconomic Pathways (SSP2-4.5 and SSP3-7.0). High-resolution regional climate simulations at 0.11° spatial resolution were generated using the COSMO-CLM model driven by EC-Earth3-Veg to capture localized climatic processes with enhanced fidelity. By deriving bioclimatic indices from these simulations, the analysis captures the spatial and temporal variability of climate-driven ecological patterns under divergent future scenarios. Model evaluation against ERA5-Land reanalysis data indicates that COSMO-CLM reasonably captures the broad spatial patterns of temperature and precipitation across Türkiye, along with the seasonal variability. However, simulations showed systematic biases in both variables, with a notable warm bias over high-elevation and interior regions and underestimation of precipitation along the Black Sea coast. Hence, simulated variables are needed for bias correction prior to computation of the bioclimatic indices. The application of QDM significantly improved the alignment with observed data. Similarly, Önol and Ünal [30] demonstrated that the simulations of the regional climate model RegCM3 at 30 km resolution exhibit a slight warm bias over Türkiye with respect to CRU gridded observations. Additionally, ref. [47] indicated that MPI-ESM-LR-driven COSMO-CLM simulations systematically yield higher annual mean temperatures than those obtained from meteorological observation data over northwestern Türkiye. The analysis of three GCMs—ECHAM5, CCSM3, and HadCM3—driven by RegCM3 outputs for transitional seasons reveals a slight cold bias over Turkey [48]. Conversely, Önol and Ünal [30] found that temperatures during the transition seasons are largely consistent with CRU observations. The results highlight that accurate regional validation is indispensable, especially in areas with complex topography where model performance diverges significantly from observations.

Analysis of future climate trends under SSP2-4.5 and SSP3-7.0 demonstrates a consistent signal over Türkiye, with varying temporal responses between scenarios. While SSP3-7.0 is characterized by a steady intensification of warming, SSP2-4.5 exhibits a wavy pattern with alternating periods of acceleration and stabilization. On the other hand, precipitation trends diverge more significantly. SSP3-7.0 projects progressive drying, becoming stronger especially after 2060, while SSP2-4.5 projects a sharp early-century drying followed by partial recovery. Both scenarios identify Eastern Anatolia as a climate sensitivity hotspot, reflecting the influence of elevation-dependent warming and shifting precipitation regimes. Even though there are differences in emission pathways, both SSP scenarios meet on the expansion of arid and semi-arid conditions, especially around the central and southeastern regions of Türkiye. The marked spatial and temporal variability in precipitation response clearly stress the importance of incorporating non-linear climate dynamics when assessing future water availability and ecosystem resilience.

The projected evolution of bioclimatic variables indicates a reasonable and spatially differentiated response to climate change across Türkiye. Bioclimatic indicators reveal additional ecological implications: daily temperature ranges remain narrow in coastal regions and wide in inland highlands, while temperature stability (isothermality) and seasonality show distinct regional patterns. Projections for both scenarios show asymmetrical seasonal warming indicating accelerated summer warming. This is driven by the fact that Türkiye is experiencing strong land–atmosphere feedbacks, altered atmospheric circulation, and land–sea thermal contrasts. Hot, dry summers limit soil moisture, while earlier spring drying and snowmelt increase sensible heating and summer solar absorption. Slower warming of the Mediterranean Sea amplifies land–sea pressure gradients, reinforcing continental heating. Weaker mid-latitude westerlies and more persistent subtropical highs reduce cool Atlantic air intrusions, prolonging heatwaves [49,50]. Reduced inland humidity further limits evaporative cooling, making the region a pronounced summer warming hotspot. Particularly under SSP3-7.0, the variations in bioclimatic indicators correspond to elevated annual and seasonal temperatures, reduced cold extremes, and increased temperature seasonality, especially in inland and high-altitude regions. Concurrently, precipitation-related indices point to increasing wetness along the eastern Black Sea coast, but drying trends and more irregular moisture regimes dominate most other regions.

Shifts in seasonal precipitation timing are also observed, with coastal areas experiencing delayed onset and mountainous regions showing earlier shifts. Delays in the onset of the wettest quarter are featured along coastal areas, particularly the Aegean region, while significant advances up to four months are observed in high-elevation and continental regions such as the Anatolian plateau. These shifts indicate a westward expansion of the winter-dominant regime and suggest increased climate sensitivity in mountainous areas. Furthermore, our findings suggest an alignment between the onset month of the wettest quarter and trends in the mean temperature of wettest quarter, proving the interconnected nature of temperature and seasonal precipitation changes. In regions where precipitation shifts toward warmer or cooler months, the corresponding mean temperature of the wettest quarter increases or decreases, confirming the projected reorganization of the seasonal hydroclimatic conditions. These findings support the need to consider quarterly bioclimatic indices when assessing regional climate impacts, particularly in ecologically and hydrologically sensitive zones. An asymmetric pattern of seasonal precipitation shifts is observed, with minor delays along coastal areas and more pronounced, multi-month advances in mountainous regions, supporting previous findings that high-elevation areas are more sensitive to future climate change [7,14].

Our results highlight the importance of integrating bioclimatic indices in regional climate assessments to better understand the ecological and agricultural impacts of climate change. While COSMO-CLM simulations at high spatial resolutions provide improved representation of mesoscale processes compared to the ESMs, and this study takes into account two alternative climate scenarios, the analysis is based on regional climate model simulations driven by a single ESM. This may limit the representations of the range of internal climate variability and uncertainty. Nonetheless, the study provides critical information into the spatiotemporal trajectories of climate-driven ecological shifts in Türkiye. The results emphasize the need for scenario-informed, regionally tailored adaptation strategies, particularly considering that Türkiye’s exposure to heat and drought intensifies toward the end of the century.

To further investigate regional climate diversity, a k-means clustering analysis [51] was performed using combined temperature and precipitation data. The analysis identifies four stable sub-climate types that evolve over time. Cluster 1 represents the coldest and moderately wet regions, covering mainly Eastern Anatolia and the inland Black Sea, with a projected warming from 5.56 °C to 10.51 °C and a decrease in precipitation. Cluster 2, dominating Central Anatolia, shows a warm, moderately wet profile and expands significantly in area throughout the century. Cluster 3 includes areas with the highest precipitation, particularly around the Black Sea and parts of the Aegean and Mediterranean, but its extent diminishes over time. Cluster 4, the hottest and driest, initially covers the southern and western regions but retreats in later periods, especially under SSP3-7.0. Our findings indicate a consistent northward expansion of arid and semi-arid zones and a contraction of temperate, humid regions. Cluster 2, representing warm continental interiors, becomes increasingly dominant in time and start to replace more temperate and humid climates. The area designated to Clusters 3 and 4, which currently exhibit higher precipitation and more moderated temperatures, decrease significantly. This climate transition reflects not only rising temperatures but also changes in precipitation timing and seasonality, with Clusters 1 and 2 associated with earlier onset of the wettest season, and Cluster 4 with delayed or unchanged timing.

A parallel clustering analysis under SSP2-4.5 identifies similar sub-climate types, though with less drastic spatial changes. Notably, the hot–arid cluster expands northward in both scenarios, while temperate zones contract, especially in Central Anatolia. All clusters exhibit warming of approximately 3–4 °C and varying precipitation changes by the end of the century. When compared with seasonal precipitation timing shifts, Clusters 1 and 2 align with regions experiencing significant negative timing shifts, while Cluster 4 corresponds to zones with neutral or positive shifts. These results underscore the potential reorganization of Türkiye’s climatic landscape and emphasize the urgency of regional-scale climate adaptation.

It should be noted that this study relies on regional climate model simulations driven by a single global model (EC-Earth3-Veg). While this allows a detailed assessment of this specific model’s projections for Türkiye, it may limit the representation of the full range of internal climate variability and associated uncertainties. To partly address this limitation, we applied two different emission scenarios, providing insights into the potential range of future climate responses. Multi-model ensemble approaches using additional well-performing downscaled CMIP6 models could further enhance the robustness of the results.

Overall, the findings underscore a significant transition toward warmer and drier conditions in Türkiye, particularly in Mediterranean and mountainous areas, aligning with broader climate change hotspot projections. These results provide a scientific basis for informing national climate adaptation policies, biodiversity conservation, and sustainable land-use strategies in the face of accelerating climate change. Despite the fact that our study does not explicitly model distributions of species, the results highlight areas where ecological conditions may shift. This information can serve as a basis for future species-specific assessments.