δ-MedBioclim: A New Dataset Bridging Current and Projected Bioclimatic Variables for the Euro-Mediterranean Region

Abstract

This data descriptor presents δ-MedBioclim, a newly developed dataset for the Euro-Mediterranean region. This dataset applies the delta-change method by comparing the values of 25 General Circulation Models (GCMs) for the reference period (1981–2010) with their projections for future periods (2026–2050, 2051–2075, and 2076–2100) under the SSP1-RCP2.6, SSP2-RCP4.5, and SSP5-RCP8.5 scenarios. These anomalies are added to two pre-existing datasets, ERA5-Land and CHELSA, yielding resolutions of 0.1° and 0.01°, respectively. Additionally, this manuscript provides a ranking of GCMs for each major river basin within the study area to guide model selection. δ-MedBioclim includes, for all the aforementioned scenarios, monthly mean temperature, total monthly precipitation, and 23 bioclimatic variables, including 9 (biorm1 to biorm9) from the Worldwide Bioclimatic Classification System (WBCS) that are not available in other databases. It also provides two bioclimatic classifications: Köppen–Geiger and WBCS. This dataset is expected to be a valuable resource for modeling the distribution of Mediterranean species and habitats, which are highly affected by climate change.

Dataset: http://dx.doi.org/10.18002/10612/24418.

Dataset License: NC-ND 4.0 International

1. Summary

The widespread availability of global climate databases does not exempt them from critical issues, such as the poor representation of regional climatology and the limited variety of General Circulation Models (GCMs) provided. These shortcomings are particularly problematic in regions highly affected by climate change, such as the European Mediterranean region.

With these challenges in mind, a new dataset, δ-MedBioclim, has been developed. Based on 25 GCMs obtained from the ESGF-LLNL node and interpolated to a 0.1° grid using Empirical Bayesian Kriging, temperature and precipitation anomalies were computed between the reference period (1981–2010) and the future periods (2026–2050, 2051–2075, and 2076–2100) under the SSP1-RCP2.6, SSP2-RCP4.5, and SSP5-RCP8.5 scenarios. Using the delta-change method, these anomalies were then applied to two well-established databases, the ERA5-Land (0.1° resolution) and CHELSA (0.01° resolution).

To complement δ-MedBioclim, in addition to temperature and precipitation values, the dataset provides 23 bioclimatic variables and two bioclimatic classifications (Köppen–Geiger and Rivas-Martínez) for each temporal period and scenario.

δ-MedBioclim is structured in a hierarchical system of folders and subfolders, ultimately containing spatial raster files (TIF format) in ETRS89 projection. Additionally, a ranking of recommended GCMs for each major river basin within the study area is provided, enabling users to make informed decisions regarding model selection.

δ-MedBioclim introduces several key innovations as a dataset: (i) it expands the range of available GCMs to 25, surpassing other databases; (ii) it employs a computationally efficient and straightforward method (delta-change) to generate projections; (iii) it includes bioclimatic variables (biorm1 to biorm9) and bioclimatic classifications (Köppen–Geiger and Rivas-Martínez) that are not available in other datasets; and (iv) it is developed for a region that is highly vulnerable to climate change yet remains underexplored in climate research. δ-MedBioclim is considered a highly applicable dataset for modelling the distribution and ecological niche of Mediterranean taxa and habitats.

2. Background

There is currently a vast number of high-resolution databases (with a spatial resolution below 1 km²) that are highly useful for ecological niche modeling [1,2,3,4,5]. These applications are based on a bioclimatic approach, which seeks to establish a relationship between climate and the distribution of living organisms [6,7]. However, it is important to acknowledge that global databases may poorly represent regional climatology, potentially affecting species distribution predictions [8]. This issue becomes even more critical in future projections, as they must account not only for the inherent uncertainty of unverifiable outcomes [9] but also for the spatial scale mismatch between climate models (general circulation models operating at scales of hundreds of kilometers) and ecological models (which function at much finer resolutions, often below one kilometer) [10].

For this reason, in recent years, various approaches have led to the development of numerous regional datasets that provide bioclimatic data that more accurately reflect local climatic conditions. Some of these studies aim to refine global projections by incorporating observational data collected in the field [11,12,13]. This approach allows for very high spatial resolutions, reaching up to 40 m, and reveals significant differences compared to global datasets, particularly in areas with complex topography [12]. However, a major drawback of this method is the need for a dense network of well-distributed meteorological stations across the region [14]. Other approaches rely on dynamic or statistical downscaling techniques or leverage subproducts derived from these methods to generate new bioclimatic datasets [15,16]. A notable example of this strategy is the recently developed BioVars dataset [17], which utilizes multiple RCMs from EURO-CORDEX to produce updated climatologies within that domain. Finally, many datasets aim to integrate global datasets at a regional scale by enhancing them with interpolated station-based products [18,19]. This approach results in “refined” versions of these datasets, which may not necessarily have higher resolution but offer improved representation of local climatological conditions.

In addition to challenges in accurately representing local climatology, many global datasets suffer from limited variability in future projection realizations. For instance, the latest version of CHELSA [1] only includes five General Circulation Models (GCMs): GFDL-ESM4, IPSL-CM6A, MPI-ESM1-2-HR, MRI-ESM2-0, and UKESM1-0-LL. Moreover, some of these models—such as IPSL and UKESM in their CMIP6 versions—are not recommended for certain regions, including Europe [20]. In contrast, WorldClim [2] offers a broader selection of GCMs, but despite the extensive range available within the WCRP, it includes only 14 models. It has been stated that increasing the number of GCMs used in a Multi-Model Ensemble (MME) reduces the associated uncertainty, enhances the representation of extreme values, and improves the alignment of predictions with observed data [21].

δ-MedBioclim aims to generate high-resolution future projections for a broad range of General Circulation Models (25 GCMs) in the highly climate-vulnerable region of the Euro-Mediterranean [22]. This is achieved by applying the delta-change method, a straightforward approach with low computational demands. As such, this dataset aligns with the urgent need for research that further advances the understanding of climate dynamics in this critically affected region [23].

3. Data Description

3.1. Regional Setting

The study area in which δ-MedBioclim has been developed corresponds to the Euro-Mediterranean Region (EMR). This area encompasses the European countries (20) located along the Mediterranean Sea and hydrologically connected to it. Figure 1 provides a physical map that includes the major cities and river basins.

3.2. Dataset Structure

Figure 2 presents a schematic representation of the folder and subfolder hierarchy of the δ-MedBioclim dataset.

δ-MedBioclim contains the following hierarchical levels of organization:

• Timeframes: At the highest folder level, the dataset is organized by time periods, with a primary folder for the reference period (1981–2010) and three additional folders corresponding to the projected time periods (one for each: 2026–2050, 2051–2075, and 2076–2100).
• At this same level, there is also a folder (shp) containing a vector file delineating the boundaries of the study area (Euro-Mediterranean region).
• Emission scenarios: At the next folder level, there are three subfolders corresponding to the selected emission scenarios (SSP1-RCP2.6, SSP2-RCP4.5, and SSP5-RCP8.5).
• Variables: Each of these “emission scenarios folders” contains four subdirectories: (i) “bio”, which includes bioclimatic parameters and variables (bio1 to bio19 and biorm1 to biorm9); (ii) “climatologies”, which stores bioclimatic classifications according to the Köppen–Geiger Classification System (KGCS) and the Worldwide Bioclimatic Classification System (WBCS); (iii) “pr”, which contains total monthly precipitation values; and (iv) “tas”, which includes mean monthly temperature values.
• GCMs: These values have been computed for the 25 selected General Circulation Models (GCMs) as well as for the Multi-Model Ensemble generated using the Random Forest approach (RF-MME). Consequently, the four subdirectories mentioned above contain 26 folders (25 GCMs + 1 RF-MME).
• Delta change method: The lowest hierarchical level, represented by the folders “d_CHELSA”, “d_E5L”, and “raw_data”, corresponds to the reference dataset used in the delta change method: CHELSA and ERA5 Land for the first two, and none for the last.
• TIF files: These files are spatial rasters in TIF format with the variable values and in the ETRS89 (EPSG:4258) projection. The file naming structure has been designed to encapsulate all the aforementioned information, ensuring that files can be used independently of their original directory structure. For instance, the file “CH_tas_2051-2075-ssp2-rcp45_UKES-M1-1-LL_01.tif” indicates, separated by “_”, the delta change reference dataset (CH = CHELSA), the variable (tas), the timeframe (2051–2075), the emission scenario (ssp2-rcp45), the GCM (UKES-M1-1-LL) and the month of the year (01 = January).

4. Methods

4.1. General Circulation Models Data

Monthly mean temperature (tas) and monthly total precipitation (pr) simulated by 25 CMIP6 General Circulation Models (GCMs) were obtained from the repository of the World Climate Research Program (WCRP) and its ESGF-LNLL node: https://esgf-node.llnl.gov/projects/cmip6/ (accessed on 16 December 2022) (

Table 1. Modeling center and acronym of the 25 General Circulation Models (GCMs) available in δ-MedBioclim.

GCM Number	GCM Acronym	Modelling Center
1	ACCESS-CM2	Commonwealth Scientific and Industrial Research Organization
2	ACCESS-ESM1-5
3	CanESM5	Canadian Centre for Climate Modelling and Analysis
4	CanESM5-CanOE
5	BCC-CSM2-MR	Beijing Climate Center
6	CAS-ESM2-0	Institute of Atmospheric Physics, Chinese Academy of Sciences
7	FGOALS-f3-L
8	FIO-ESM2-0	The First Institution of Oceanography
9	CNRM-CM6-1	Centre National de Recherches Météorologiques
10	CNRM-ESM2-1
11	IPSL-CM6A-LR	Institut Pierre-Simon Laplace/Centre National de Recherche Scientifique
12	MPI-ESM1-2-LR	Max-Planck-Institut fuer Meteorologie, Deutsches Klimarechenzentrum
13	MPI-ESM1-2-HR
14	CMCC-ESM2	Centro Euro-Mediterraneo sui Cambiamenti Climatici
15	MIROC6	National Institute for Environmental Studies, Japan Agency for Marine-Earth Science and Technology
16	MIROC-ES2L
17	MRI-ESM2-0	Meteorological Research Institute
18	INM-CM5-0	Russian Academy of Sciences, Institute of Numerical Mathematics
19	KACE-1-0-G	National Institute of Meteorological Sciences-Korea Met. Administration
20	HadGEM-GC31-LL	Met Office Hadley Centre
21	UKES-M1-1-LL
22	GFDL-ESM4	Geophysical Fluid Dynamics Laboratory/NOAA
23	GISS-E2-1-G	NASA Goddard Institute for Space Studies
24	GISS-E2-2-G
25	MCM-UA-1-0	University of Arizona—Department of Geosciences

).

Only the first ensemble run (r1i1p1) from the historical experiment (hist) and from three Shared Socioeconomic Pathways (SSPs: ssp126, ssp245, and ssp585) experiments was employed. The hist period was restricted to 1981–2010. Regarding the future periods, they were divided into a near-term future (2026–2050), a mid-term future (2051–2075), and a long-term future (2076–2100).

Given that the employed GCMs have different native spatial resolutions, an interpolation method was applied to standardize all of them onto a common 0.1° grid. To achieve this, Empirical Bayesian Kriging was implemented, as it has demonstrated minimal data distortion in the downscaling of GCMs [25,26]. ArcGIS Pro 3.1.0. and its extension “Geostatistical Analyst” was used in this regard [25].

However, it is essential to acknowledge that each GCM exhibits inherent biases and associated uncertainties. The most immediate of these pertains to the potential lack of correlation between observational data and simulated outputs—that is, a limited fidelity to empirical reality. From a more epistemological standpoint, it may be argued that such projections are unfalsifiable, as forecasts concerning future conditions are, by definition, unverifiable in the present [9].

To mitigate these deviations, it is common practice to generate consensus models, commonly referred to as Multi-Model Ensembles (MMEs) [21,26,27]. A widely used approach due to its relative simplicity and strong performance is the Random Forest-based Multi-Model Ensemble (RF-MME) [28,29,30]. In the present dataset, alongside the individual outputs of each GCM, an ensemble derived from this technique is included, incorporating all 25 GCMs available in δ-MedBioclim. However, issues like the interoperability of models may be taken into account, as it may conflict with their independence, a factor that should be considered when constructing MMEs [31].

In conclusion, the results obtained using the present dataset—or others of a similar nature—should be regarded as hypotheses, i.e., as plausible scenarios of change accompanied by associated uncertainty, rather than as absolute or deterministic outcomes.

4.2. Delta Change Method

To obtain corrected future projections, the delta change method was applied. This method involves calculating the anomalies between the projected value of the variable and its current value, which are then contrasted with a reference value [32,33]. In the case of temperature (tas), the change fraction method was applied [Equation (1)], while for precipitation (pr), relative anomalies and multiplication were used [Equation (2)], as demonstrated in the following equations [34]:

$${tas}_{corr}={(tas}_{proj}-{tas}_{curr})+{tas}_{ref}$$

(1)

$${pr}_{corr}={\displaystyle \frac{{pr}_{proj}}{{pr}_{curr}}}·{pr}_{ref}$$

(2)

where tas_corr and pr_corr represent the bias-corrected future values, tas_proj and pr_proj correspond to the raw future projections from the GCM, tas_curr and pr_curr denote the GCM-derived values for the historical period (1981–2010), and tas_ref and pr_ref refer to the reference values for the same period.

Through the application of the delta change method, it is also possible to rescale and achieve finer spatial resolutions. Thus, this process was applied using two reference datasets:

• ERA5 Land [35]: With the same resolution as that achieved through the interpolation of the GCMs (0.1°).
• CHELSA [1]: With a finer resolution (0.01°).
• The use of the delta change method to obtain derived datasets with a finer resolution than the original dataset has already been implemented in previous research [19,36,37].
• Although the delta change method has been positively assessed, particularly for Southern Europe [32], several limitations must be considered, especially with regard to precipitation estimates. Previous studies have shown that in particularly arid regions, such as those encompassed by the present dataset, precipitation values may be substantially overestimated [37]. Therefore, these values should be appropriately corrected before being applied in subsequent analyses. Similarly, by applying a uniform change factor to a given area, the delta change method may lead to a loss of extreme values [38]. Consequently, studies focusing on climate extremes are advised to employ bias correction techniques based on quantiles (e.g., above the 95th or 99th percentile) rather than relying solely on change factor methods.

4.3. Bioclimatic Variables

In addition to the climatic variables of mean monthly temperature and total monthly precipitation, the necessary calculations were performed to derive key bioclimatic indices [2]: bio1 to bio19, with some exclusions (bio2, bio3, bio5, bio6, and bio7) due to data unavailability.

Furthermore, additional parameters and bioclimatic indices from the Worldwide Bioclimatic Classification System (WBCS) [6] were computed and incorporated into the dataset. This classification system is particularly noteworthy for its strong correlation with vegetation and plant formations [39,40,41,42], and its indices have been successfully integrated into species distribution modeling, yielding promising results [43,44,45,46]. The acronyms, calculations, and definitions of the indices and parameters included in δ-MedBioclim can be found in

Table 2. Bioclimatic parameters and indexes included in δ-MedBioclim. Abbreviations: tas_i—monthly average temperature, sd—standard deviation; pr_i—monthly total precipitation; tasmax—monthly average temperature of the warmest month; tasmin—monthly average temperature of the coldest month; tasp_i—monthly average temperature if it is above 0 °C (if not tasp_i = 0); prp_i—monthly total precipitation of a month where tasp_i > 0 (if not prp_i = 0).

Acronym	Definition	Calculation
bio1	Mean annual temperature	${\displaystyle \raisebox{1ex}{${\sum }_{i=1}^{12}{tas}_i$}\!\left/ \!\raisebox{-1ex}{$12$}\right.}$
bio4	Temperature seasonality (i.e., standard deviation of average monthly temperature)	$sd[{tas}_1,\left(\dots \right),{tas}_{12}]$
bio8	Mean temperature of the wettest quarter of the year
bio9	Mean temperature of the driest quarter of the year
bio10	Mean temperature of the warmest quarter of the year
bio11	Mean temperature of the coldest quarter of the year
bio12	Total annual precipitation	${\displaystyle \sum _{i=1}^{12}}{pr}_i$
bio13	Total precipitation of the wettest month of the year
bio14	Total precipitation of the driest month of the year
bio15	Precipitation seasonality (i.e., standard deviation of monthly total precipitation)	$sd[{pr}_1,\left(\dots \right),{pr}_{12}]/D$ $D=1+(bio12/12)$
bio16	Total precipitation of the wettest quarter of the year
bio17	Total precipitation of the driest quarter of the year
bio18	Total precipitation of the warmest quarter of the year
bio19	Total precipitation of the coldest quarter of the year
biorm1 (=Ic)	Continentality index (i.e., difference between the maximum and the minimum monthly average temperature)	$tasmax-tasmin$
biorm2 (=It/Itc)	Compensated thermicity index (or cold index) (1,2)	$2·bio1·10$
biorm3 (=Tp)	Positive annual temperature (i.e., sum of all temperatures above 0 °C)	$10{\displaystyle \sum _{i=1}^{12}}{tasp}_i$
biorm4 (=Pp)	Positive annual precipitation (i.e., sum of precipitation for each month where temperature is above 0 °C)	${\displaystyle \sum _{i=1}^{12}}{prp}_i$
biorm5 (=Io)	Annual ombrothermic index (i.e., ratio between positive precipitation and positive temperature)	${\displaystyle \raisebox{1ex}{$biorm4·10$}\!\left/ \!\raisebox{-1ex}{$biorm3$}\right.}$
biorm6 (=Ios1)	Ombrothermic index of the warmest month
biorm7 (=Ios2)	Ombrothermic index of the two consecutive warmest summer months (3)
biorm8 (=Ios3)	Ombrothermic index of the summer quarter
biorm9 (=Ios4)	Ombrothermic index of the summer quarter and the preceding month

⁽¹⁾ An approximation using bio1 was implemented, as we cannot obtain maximum and minimum temperatures. The actual expression of biorm2 is as follows: $biorm2=10·(bio1+M+m)$, where M denotes the annual maximum temperature of the coldest month and m represents the annual minimum temperature of the coldest month. ⁽²⁾ biorm2 has to be corrected depending on the continentality (biorm1) values. (a) If biorm1 $\le 8\to biorm{2}^{\prime }=biorm2-10·(8-biorm1)$; (b) If 18 $\le$ biorm1 $\le 21\to biorm{2}^{\prime }=biorm2+5·(biorm1-18)$; (c) If 21 $\le$ biorm1 $\le 28\to biorm{2}^{\prime }=biorm2+15+15·(biorm1-21)$; (d) If 28 $\le$ biorm1 $\le 46\to biorm{2}^{\prime }=biorm2+15+105+25·(biorm1-28)$; (e) If biorm1 $\ge 46\to biorm{2}^{\prime }=biorm2+15+105+450+30·(biorm1-46)$. ⁽³⁾ Usually, July and August (in the northern hemisphere).

.

4.4. Climatologies

Finally, as a complement to the bioclimatic parameters and variables, the climatic characterization of the study area was conducted following the Köppen–Geiger Climate Classification System (KGCS) [47] and the Worldwide Bioclimatic Classification System (WBCS) [6]. This characterization was performed for the reference period (1981–2010) as well as for three future time periods (2026–2050, 2051–2075, and 2076–2100) under the three emission scenarios considered (SSP1-RCP2.6, SSP2-RCP4.5, and SSP5-RCP8.5).

A summary of the KGCS climate classes (B, C, and E) and the macrobioclimates (Mediterranean and Temperate) identified within the EMR can be found in

Table 3. Arid (B), temperate (C) and polar (E) conditions of the Köppen–Geiger Climate Classification System. Abbreviations: MAT—mean annual temperature (°C); T_cold—temperature of the coldest month (°C); T_hot—temperature of the warmest month (°C); T_m10—number of months with mean temperature > 10 °C; MAP—mean annual precipitation (mm); P_th—2∙MAT if >70% of precipitation falls in winter, 2∙MAT + 28 if >70% of precipitation fall in summer, 2∙MAT + 14 otherwise; P_swet—precipitation of the summer wettest month; P_wwet—precipitation of the winter wettest month; P_sdry—precipitation of the summer dryest month; P_wdry—precipitation of the winter dryest month.

Class/Climate	Type	Subtype
Arid (B) $MAP<10P_{th}$	Desert (Bw) $MAP<5P_{th}$	Hot (h) $MAT>18$
	Steppe (Bh) $MAP>5P_{th}$	Cold (k) $MAT<18$
Temperate (C) $T_{hot}>10\&$ $0<T_{cold}<18$	Dry summer (Cs) $P_{sdry}<40\&P_{sdry}<3P_{wwet}$	Hot summer (a) $T_{hot}>22$
	Dry winter (Cw) $P_{wdry}<10P_{swet}$	Warm summer (b) $T_{m10}\ge 4$
	Without dry season (Cf) $NotCsorCw$	Cold summer (c) $4>T_{m10}>0$
Polar (E) $T_{hot}<10$	Tundra (Et) $T_{hot}>0$ Frost (Ef) $T_{hot}\le 0$

and

Table 4. Mediterranean and temperate conditions in the Worldwide Bioclimatic Classification System. Abbreviations: T—mean annual temperature; m—mean minimum temperature of the coldest month; M—mean maximum temperature of the coldest month; It/Itc—(compensated) thermicity index (10(T + M + m)); Ic—index of continentality (difference between mean annual maximum temperature and mean annual minimum temperature); Io—annual ombrothermic index (ratio between the sum of monthly temperatures exceeding 0 °C and the precipitation of those months); Ios₂—ombrothermic index of the two summer warmest months, Ios₁—ombrothermic index of the warmest month, Ios₃—ombrothermic index of the two summer warmest months and the preceding month, Ios₄—ombrothermic index of the two summer warmest months and the two preceding months.

Macrobioclimate	Bioclimate
Mediterranean (M.) (1) If ((23° to 52° N-S), ${Ios}_2<2.0$) If ((23° to 36° N-S), ${Ios}_2<2.0$ & ($T<25.0$ & $m<10.0$) or ($T<25.0$ & $It/Itc<580$) or ($m<10.0$ & $It/Itc<580$))	M. desertic oceanic (medo) $Ic\le 21.0\&0.2<Io<1$.0
	M. desertic continental (medc) $Ic\ge 21.0\&0.2<Io<1$.0
	M. xeric oceanic (mexo) $Ic\le 21.0\&1.0<Io<2$.0
	M. xeric continental (mexc) $Ic\ge 21.0\&1.0<Io<2.0$
	M. pluviseasonal oceanic (mepo) $Ic\le 21.0\&Io>2.0$
	M. pluviseasonal continental (mepc) $Ic\ge 21.0\&Io>2.0$
Temperate (T.) If ((23° to 66° N- 54° S), ${Ios}_2>2.0$) If ((23° to 36° N-S), ${Ios}_2>2.0$ & $T<21.0$ & $M<18.0$ & $It/Itc<470$)	T. hyperoceanic (teho) $Ic\le 11.0\&Io>3.6$
	T. oceanic (teoc) $11.0<Ic<21.0\&Io>3.6$
	T. continental (teco) $Ic\ge 21.0\&Io>3.6$
	T. xeric (texe) $Ic\le 11.0\&Io\le 3.6$

⁽¹⁾ If in any summer month Ios₁ < 2.8, the macrobioclimate cannot be Mediterranean, and it will be Temperate with a submediterranean bioclimatic variant. It will also happen if the following condition is fulfilled: ((Io > 12) and (Ios₂ > 1.9) and (Ios₃ > 2.0) and (Ios₄ > 2.0)) or ((11 < Io < 10) and (Ios₂ > 1.8) and (Ios₃ > 1.9) and (Ios₄ > 2.0)) or ((10 < Io < 9) and (Ios₂ > 1.8) and (Ios₃ > 1.9) and (Ios₄ > 2.0)) or ((9 < Io < 8) and (Ios₂ > 1.7) and (Ios₃ > 1.9) and (Ios₄ > 2.0)) or (8 < Io < 7 and (Ios₂ > 1.5) and (Ios₃ > 1.8) and (Ios₄ > 2.0)) or (7 < Io < 6 and (Ios₂ > 1.4) and (Ios₃ > 1.8) and (Ios₄ > 2.0)) or ((6 < Io < 4.8 and (Ios₂ > 1.3) and (Ios₃ > 1.8) and (Ios₄ > 2.0)) or (4.8 < Io < 3.6 and (Ios₂ >1.2) and (Ios₃ > 1.7) and (Ios₄ > 2.0)) or (3.6 < Io < 2.8 and (Ios₂ > 1.1) and (Ios₃ > 1.7) and (Ios₄ > 2.0)) or (2.8 < Io < 2 and (Ios₂ > 0.9) and (Ios₃ > 1.7) and (Ios₄ > 2.0)).

, respectively. For further details, readers are encouraged to consult the original publications in their entirety.

5. User Notes

GCMs Ranking

Below, a ranking of the five General Circulation Models (GCMs) demonstrating the best performance is provided for each of the major river basins in the study area (

Table 5. Best-performing GCM for each major river basin (Level 5 [24]) within the study area. The table presents the top-performing model for temperature, precipitation, and the overall ranking considering both variables. The numerical identifiers (ID) corresponding to each GCM can be found in Table 1.

Country	River Basin Name—Code	Temperature (Numerical ID)	Precipitation (Numerical ID)	Overall
Austria/Hungary	North Danube—2040539930	16, 19, 20, 11, 7	18, 23, 8, 9, 10	16, 8, 19, 5, 9
Austria/Slovenia	Drava—2040540100	13, 2, 10, 18, 24	22, 25, 14, 18, 6	22, 25, 18, 24, 13
Bosnia/Croatia	Sava—2040555780	24, 22, 8, 10, 9	24, 18, 22, 4, 6	24, 22, 8, 9, 10
France	Garonne—2040020320	12, 14, 1, 15, 17	4, 22, 10, 6, 2	12, 17, 22, 4, 23
France	Loire—2040021030	24, 17, 8, 10, 9	18, 16, 4, 11, 6	8, 10, 11, 16, 6
France	Brittany—2040021040	12, 22, 1, 10, 5	14, 4, 9, 22, 8	22, 1, 12, 15, 4
France	Seine—2040022150	25, 17, 3, 15, 4	18, 11, 6, 16, 4	16, 18, 4, 25, 6
France	Artois–Le Meuse—2040022160	15, 12, 1, 20, 3	5, 16, 4, 22, 25	15, 16, 22, 5, 12
France/Switzerland	Rhin—2040023010	19, 8, 24, 13, 3	5, 16, 18, 10, 3	8, 10, 16, 19, 3
Greece/Albania	Ionian–Adriatic—2040009230	22, 9, 24, 8, 11	22, 4, 18, 12, 6	22, 24, 6, 18, 9
Greece/Turkey	Greek islands—2040045150	7, 18, 25, 16, 9	10, 20, 7, 12, 17	7, 20, 10, 25, 16
Italy	Po–East Apennines—2040012730	12, 22, 24, 8, 14	17, 4, 20, 22, 12	12, 22, 17, 24, 9
Italy	Tyrrhenian—2040014550	5, 9, 4, 8, 24	14, 13, 5, 19, 10	5, 13, 4, 17, 9
Italy	Sicily—2040046500	12, 3, 22, 17, 24	20, 18, 1, 19, 6	12, 24, 18, 1, 17
Italy/France	Corsica–Sardinia—2040047500	8, 24, 9, 12, 3	13, 19, 14, 7, 22	22, 24, 13, 23, 3
Portugal	Mondego—2040018470	10, 20, 13, 24, 16	3, 16, 4, 11, 2	16, 10, 3, 19, 23
Portugal	Sado—2040018850	16, 20, 22, 15, 18	4, 16, 18, 22, 15	16, 22, 18, 15, 4
Romania	Prut—2040008490	20, 16, 3, 19, 24	25, 15, 7, 18, 16	16, 18, 25, 3, 19
Romania	Siret—2040543090	12, 14, 24, 22, 17	22, 12, 24, 1, 13	12, 22, 24, 1, 17
Romania/Bulgaria	Black Sea—2040008500	12, 15, 18, 25, 24	18, 12, 25, 20, 3	12, 18, 25, 15, 20
Romania/Hungary	Somes–Mures—2040548500	24, 8, 2, 18, 7	18, 7, 11, 8, 14	18, 8, 7, 24, 10
Romania/Serbia	West Danube—2040543160	22, 10, 8, 19, 9	4, 22, 10, 19, 25	22, 10, 19, 11, 24
Spain	Guadalete–Barbate—2040018240	3, 12, 18, 13, 1	22, 12, 18, 4, 15	12, 22, 18, 2, 3
Spain	Guadalquivir—2040018360	12, 8, 9, 24, 22	24, 17, 14, 22, 20	24, 22, 17, 12, 14
Spain	Tinto–Odiel—2040018370	23, 22, 2, 12, 24	18, 2, 14, 17, 4	2, 23, 22, 18, 12
Spain	Balearic islands—2040048590	3, 13, 12, 23, 24	5, 23, 19, 3, 10	23, 3, 5, 13, 24
Spain/France	Ebro–Rhone—2040016230	12, 6, 24, 17, 8	4, 24, 12, 20, 17	12, 24, 17, 6, 10
Spain/Portugal	Guadiana—2040018460	20, 10, 5, 6, 24	6, 11, 3, 4, 2	6, 11, 4, 2, 20
Spain/Portugal	Tagus—2040018840	10, 6, 24, 7, 22	2, 19, 4, 14, 18	6, 10, 24, 2, 18
Spain/Portugal	Douro—2040019150	17, 24, 23, 1, 15	25, 2, 3, 15, 11	25, 15, 3, 23, 11
Spain/Portugal	Minho—2040019160	13, 14, 15, 8, 22	15, 2, 25, 6, 20	15, 14, 25, 1, 3
Turkey	Bosphorus—2040003440	22, 25, 17, 23, 24	22, 7, 16, 6, 13	22, 17, 23, 24, 6
Turkey	Sakarya—2040003550	22, 7, 10, 11, 6	25, 15, 4, 19, 6	6, 10, 22, 11, 4
Turkey	Karadeniz—2040003560	12, 15, 17, 24, 14	6, 2, 12, 3, 20	12, 15, 1, 24, 2
Turkey	Kizilirmak—2040004130	22, 7, 13, 11, 10	18, 25, 15, 4, 6	22, 10, 13, 6, 11
Turkey	Yesilirmak—2040004280	12, 15, 13, 7, 17	18, 19, 25, 15, 3	18, 15, 25, 22, 12
Turkey	Çoruh—2040005120	12, 17, 22, 10, 1	25, 6, 15, 11, 3	15, 25, 12, 1, 14
Turkey	Dicle—2040816320	23, 9, 20, 6, 8	6, 10, 14, 15, 21	6, 23, 20, 13, 10
Turkey	Firat—2040785900	7, 11, 23, 20, 24	6, 11, 16, 25, 10	11, 6, 23, 20, 25
Turkey	Van Goku—2040085990	15, 13, 9, 8, 25	6, 15, 11, 25, 22	15, 6, 25, 4, 22
Turkey/Cyprus	Aegean—2040000010	6, 18, 5, 23, 13	10, 19, 5, 25, 11	6, 5, 25, 13, 8

). In this regard, a river basin was considered as each unit of Level 5 of HydroSHEDS [24]. On the other hand, the ranking has been established based on the simulation accuracy of mean monthly temperature and total monthly precipitation, following the findings of [48,49]. It is provided with the intention of allowing the user to make an informed decision on which GCMs to use in each subregion of the study area. For more details about performance estimation or ranking obtention, readers are encouraged to consult the aforementioned publications.