ABSTRACT
Meteorological observations in Iran indicate a clear trend toward warming and drying. To assess how well current climate models capture these changes, we evaluated twenty CMIP6 General Circulation Models (GCMs) against observed spatial patterns of precipitation (Pr), minimum temperature (Tasmin), and maximum temperature (Tasmax) over 1987–2014. Model performance was quantified using the Kling–Gupta efficiency (KGE) and mean absolute error (MAE), leading to a ranking in which AWI‑CM‑1‑1‑MR and BCC‑CSM2‑MR emerged as the best performers. These two models were then used to project future climate under three Shared Socioeconomic Pathway scenarios (SSP126, SSP245, and SSP585) for the period 2060–2099. All scenarios indicate continued increases in both Tasmin and Tasmax and decreases in Pr, with the most severe changes under SSP585. AWI‑CM‑1‑1‑MR predicts a 4.1 °C rise in Tasmin, a 4.9 °C rise in Tasmax, and an annual precipitation decline of 7.1 mm, while BCC‑CSM2‑MR forecasts slightly more moderate shifts. Spatially, northern Iran faces greater temperature increases and rainfall reductions than the south. These projections underscore an urgent need for targeted adaptation measures to safeguard water resources, agriculture, human health, and other climate‑sensitive sectors in Iran.
HIGHLIGHTS
AWI-CM-1-1-MR and BCC-CSM2-MR were identified as the most optimal GCMs for predicting climate changes in Iran.
By 2060–2099, the worst-case SSP585 scenario projects a 4.1 °C increase in (Tasmin) and a 4.9 °C increase in (Tasmax).
A projected reduction of 7.1 mm in (Pr) was observed, signaling worsening drought conditions in Iran.
INTRODUCTION
Climate change, one of the most pressing environmental crises of our era, has been increasingly threatening ecosystems, societies, and the global economy (Lungarska & Chakir 2024; Zhang, T. et al. 2024). This phenomenon, which is known for a sustained change in worldwide or regional weather patterns, manifests in a variety of forms, including an increase in temperature (Tsai et al. 2024), a transformation in precipitation (Pr) patterns (Sun et al. 2024), the occurrence of extreme weather events, and an elevation in sea levels (Doorga et al. 2024). The key reason for climate change is the substantial rise in atmospheric greenhouse gas concentrations (Shah et al. 2024). It has disturbed the Earth's energy balance and led to adverse global environmental, social, and economic consequences (Ding et al. 2024). Along with the alteration in Pr patterns and temperature, the rise in greenhouse gas concentrations has accelerated the frequency, duration, and severity of climate-change-related events (Hosseini et al. 2020; Sharafati & Pezeshki 2020). It is of great importance to gain an understanding of how hydrological processes are affected by climate change and to measure the uncertainty related to the hydrological response at a regional scale (Bekele et al. 2019). The intensification of the Earth's climate dynamics has led to significant alterations in the equilibrium of the Earth's system. Consequently, the frequency and intensity of floods, heat waves, droughts, and ecosystem disturbances are increasing.
In Earth's dynamic environment, a comprehensive and accurate understanding of climate dynamics and the ability to forecast changes are essential to guarantee a sustainable future. Meanwhile, atmospheric general circulation models (GCMs), highly sophisticated computational tools, have made notable advances in this field (Xue et al. 2024). These intricate mathematical models, which simulate the physical processes that govern air and ocean circulation, provide insight into the Earth's climate system, offering a more profound comprehension of its complex dynamics. By incorporating factors such as radiation transport, heat exchange, and moisture transfer within three-dimensional grids, GCMs can simulate large-scale displacements of air and water masses (Peng et al. 2020). The simulations provide a comprehensive picture of climate patterns at the global level, which is vital for studying climate change and formulating strategies to deal with it (Illangasingha et al. 2023). While GCMs possess considerable capabilities, their performance is not without constraints. The limited spatial resolution of these models prevents comprehensive documentation of minor climatic events or regional transformations (Guo et al. 2023). However, Using representative concentration pathways, GCMs project future climate changes that provide critical insights for large-scale planning and policy decisions at national and global levels. There is no doubt that GCMs are indispensable for understanding and predicting the climate system's behavior under different greenhouse gas emission scenarios (Sun et al. 2023). Such models enable scientists to study the intricate dynamics and processes that drive climate change.
However, selecting the most suitable GCMs for a particular region represents a significant challenge, given the diverse range of existing models with varying degrees of strength and weakness (Raju & Kumar 2020). This challenge can be due to various factors. For example: (1) Inherent uncertainty in GCM models: These models are formulated based on complex mathematical equations and different assumptions, which result in an inherent uncertainty in their outcomes. The selection of an appropriate model necessitates an understanding of and ability to manage these uncertainties (Zhang, B. et al. 2024). (2) Limitations of GCM spatial resolution: GCMs are generally for global or continental scales that may not afford sufficient spatial resolution to record local climate features accurately. It is crucial to select a model with an appropriate spatial resolution for the target area (Benedict et al. 2017). (3) Variance in model performance: the performance of different GCM models in predicting weather variables, such as Pr, temperature, and wind, varies. It is, therefore, necessary to select a model that can perform well for the intended area (Hodnebrog et al. 2022). The selection of an appropriate model necessitates a comprehensive grasp of the mentioned challenges and their effective management.
Iran's diverse climatic conditions are undergoing profound transformations due to global climate change, marked by increasing temperature extremes and erratic Pr patterns, exacerbating water scarcity. The country's topography, comprising arid and semi-arid regions, renders it particularly susceptible to the impacts of weather anomalies (Abedi Sarvestani & Millar 2024). Recent studies have highlighted the necessity to address the rising frequency and intensity of droughts, which present significant challenges to the sustainability of agriculture, water resources management, and economic and social stability. The need for robust climate adaptation strategies is underscored by predictions of continued warming and Pr variability, requiring interdisciplinary research and policy integration to reduce negative impacts and strengthen resilience to cope with uncertain climate futures (Pakrooh & Kamal 2023).
This research focuses on the performance evaluation of GCM in the spatial pattern simulation of Pr, minimum temperature (Tasmin), and maximum temperature (Tasmax) across Iran for the period of 1987 to 2014. A set of GCMs from the Coupled Model Intercomparison Project Phase 6 (CMIP6) archive has been nominated according to global performance and availability. The Kling–Gupta efficiency (KGE) and mean absolute error (MAE) were employed to assess the model's performance in simulating spatial patterns of climate variables. KGE evaluates the model's ability to accurately capture the observed variability, mean, and correlation. MAE provides a measure of the average magnitude of the errors between the simulated and observed values, offering insights into the model's systematic and random errors. Iran is currently facing critical challenges of water scarcity, and climate change projections indicate that the situation will worsen (Talebi 2023). Accurate predictions of future temperature and Pr patterns are essential for developing effective water management strategies and climate adaptation programs. This study attempts to contribute to this important endeavor by providing a framework for selecting GCMs that produce reliable climate change predictions for Iran.
The area of the study
Description of various Pr zones in Iran (Modarres 2006)
Group . | Area (%) . | Remarks . |
---|---|---|
G1 | 52.04 | Semi-arid and arid, minimum yearly Pr, no summer rainfall |
G2 | 12.04 | The Highland surroundings, Pr in spring and winter |
G3 | 8.55 | Northwest and cold, Pr in the spring and little rain in the summer |
G4 | 16.31 | The Persian Gulf Margins, Pr in winter |
G5 | 4.93 | Zagros mountain range, Pr in spring and winter |
G6 | 3.09 | The Caspian Sea Margins, Pr in summer |
G7 | 2.08 | The Zagros Mountains, rainfall in Spring and Winter, Pr G7 > Pr G2 and G5 |
G8 | 0.96 | The Caspian Sea Margins, summer Pr, Pr G8 > Pr G6 |
Group . | Area (%) . | Remarks . |
---|---|---|
G1 | 52.04 | Semi-arid and arid, minimum yearly Pr, no summer rainfall |
G2 | 12.04 | The Highland surroundings, Pr in spring and winter |
G3 | 8.55 | Northwest and cold, Pr in the spring and little rain in the summer |
G4 | 16.31 | The Persian Gulf Margins, Pr in winter |
G5 | 4.93 | Zagros mountain range, Pr in spring and winter |
G6 | 3.09 | The Caspian Sea Margins, Pr in summer |
G7 | 2.08 | The Zagros Mountains, rainfall in Spring and Winter, Pr G7 > Pr G2 and G5 |
G8 | 0.96 | The Caspian Sea Margins, summer Pr, Pr G8 > Pr G6 |
List of investigated meteorological stations in eight regions of Iran
Code . | No. . | Station . | Zone . | Latitude . | Longitude . | Code . | No. . | Station . | Zone . | Latitude . | Longitude . |
---|---|---|---|---|---|---|---|---|---|---|---|
40898 | 0 | Chabahar | 1 | 25.281 | 60.651 | 40726 | 29 | Mahabad | 3 | 36.753 | 45.715 |
40856 | 1 | Zahedan | 1 | 29.472 | 60.900 | 40716 | 30 | Mianeh | 3 | 37.450 | 47.700 |
40829 | 2 | Zabol | 1 | 31.089 | 61.543 | 40811 | 31 | Ahvaz | 4 | 31.345 | 48.744 |
40854 | 3 | Bam | 1 | 28.633 | 58.383 | 40831 | 32 | Abadan | 4 | 30.377 | 48.215 |
40841 | 4 | Kerman | 1 | 30.256 | 56.963 | 40794 | 33 | Safiabaddezfool | 4 | 32.253 | 48.433 |
40851 | 5 | Sirjan | 1 | 29.867 | 55.750 | 40833 | 34 | Omidiyehaghajari | 4 | 30.743 | 49.688 |
40821 | 6 | Yazd | 1 | 31.904 | 54.290 | 40858 | 35 | Bushehr | 4 | 28.904 | 50.821 |
40809 | 7 | Birjand | 1 | 32.891 | 59.284 | 40859 | 36 | Fasa | 4 | 28.899 | 53.719 |
40785 | 8 | Kashan | 1 | 33.967 | 51.481 | 40875 | 37 | Bandarabas | 4 | 27.215 | 56.373 |
40789 | 9 | Khurvabiabanak | 1 | 33.770 | 55.082 | 40893 | 38 | Jask | 4 | 25.638 | 57.770 |
40763 | 10 | Kashmar | 1 | 35.271 | 58.473 | 40848 | 39 | Shiraz | 4 | 29.561 | 52.603 |
40757 | 11 | Semnan | 1 | 35.588 | 53.421 | 40883 | 40 | Bandarlenge | 4 | 26.528 | 54.828 |
40800 | 12 | Isfahan | 1 | 32.518 | 51.706 | 40818 | 41 | Abadeh | 4 | 31.198 | 52.616 |
40827 | 13 | Nahabandan | 1 | 31.542 | 60.035 | 40812 | 42 | Masjed Solyeman | 4 | 31.983 | 49.241 |
40745 | 14 | Mashhad | 2 | 36.236 | 59.631 | 40783 | 43 | Aligudarz | 5 | 33.408 | 49.703 |
40743 | 15 | Sabzevar | 2 | 36.207 | 57.649 | 40782 | 44 | Khorramabad | 5 | 33.439 | 48.284 |
40723 | 16 | Bojnurd | 2 | 37.533 | 57.117 | 40747 | 45 | Sanandaj | 5 | 35.254 | 47.015 |
40740 | 17 | Quchan | 2 | 37.117 | 58.450 | 40727 | 46 | Saqez | 5 | 36.222 | 46.311 |
40754 | 18 | Tehran | 2 | 35.683 | 51.433 | 40766 | 47 | Kermanshah | 5 | 34.352 | 47.153 |
40731 | 19 | Qazvin | 2 | 36.262 | 50.061 | 40736 | 48 | Babolsar | 6 | 36.699 | 52.643 |
40769 | 20 | Arak | 2 | 34.072 | 49.783 | 40732 | 49 | Nowshar | 6 | 36.661 | 51.467 |
40768 | 21 | Hamedan | 2 | 34.869 | 48.535 | 40738 | 50 | Gorgan | 6 | 36.905 | 54.414 |
40798 | 22 | Shahrekord | 2 | 32.292 | 50.839 | 40732 | 51 | Ramsar | 6 | 36.904 | 50.683 |
40706 | 23 | Tabriz | 3 | 38.122 | 46.243 | 40737 | 52 | Ghaemshahrgharakheil | 6 | 36.454 | 52.772 |
40712 | 24 | Urmia | 3 | 37.659 | 45.055 | 40780 | 53 | Ilam | 7 | 33.588 | 46.398 |
40729 | 25 | Zanjan | 3 | 36.660 | 48.522 | 40836 | 54 | Yasuj | 7 | 30.699 | 51.555 |
40708 | 26 | Ardebil | 3 | 38.218 | 48.329 | 40835 | 55 | DogonbadanBandaranzali | 7 | 30.346 | 50.819 |
40700 | 27 | Parsabadmoghan | 3 | 39.648 | 47.917 | 40718 | 56 | Bandaranzali | 8 | 37.480 | 49.458 |
40703 | 28 | Khoy | 3 | 38.558 | 44.995 | 40719 | 57 | Rasht | 8 | 37.323 | 49.624 |
Code . | No. . | Station . | Zone . | Latitude . | Longitude . | Code . | No. . | Station . | Zone . | Latitude . | Longitude . |
---|---|---|---|---|---|---|---|---|---|---|---|
40898 | 0 | Chabahar | 1 | 25.281 | 60.651 | 40726 | 29 | Mahabad | 3 | 36.753 | 45.715 |
40856 | 1 | Zahedan | 1 | 29.472 | 60.900 | 40716 | 30 | Mianeh | 3 | 37.450 | 47.700 |
40829 | 2 | Zabol | 1 | 31.089 | 61.543 | 40811 | 31 | Ahvaz | 4 | 31.345 | 48.744 |
40854 | 3 | Bam | 1 | 28.633 | 58.383 | 40831 | 32 | Abadan | 4 | 30.377 | 48.215 |
40841 | 4 | Kerman | 1 | 30.256 | 56.963 | 40794 | 33 | Safiabaddezfool | 4 | 32.253 | 48.433 |
40851 | 5 | Sirjan | 1 | 29.867 | 55.750 | 40833 | 34 | Omidiyehaghajari | 4 | 30.743 | 49.688 |
40821 | 6 | Yazd | 1 | 31.904 | 54.290 | 40858 | 35 | Bushehr | 4 | 28.904 | 50.821 |
40809 | 7 | Birjand | 1 | 32.891 | 59.284 | 40859 | 36 | Fasa | 4 | 28.899 | 53.719 |
40785 | 8 | Kashan | 1 | 33.967 | 51.481 | 40875 | 37 | Bandarabas | 4 | 27.215 | 56.373 |
40789 | 9 | Khurvabiabanak | 1 | 33.770 | 55.082 | 40893 | 38 | Jask | 4 | 25.638 | 57.770 |
40763 | 10 | Kashmar | 1 | 35.271 | 58.473 | 40848 | 39 | Shiraz | 4 | 29.561 | 52.603 |
40757 | 11 | Semnan | 1 | 35.588 | 53.421 | 40883 | 40 | Bandarlenge | 4 | 26.528 | 54.828 |
40800 | 12 | Isfahan | 1 | 32.518 | 51.706 | 40818 | 41 | Abadeh | 4 | 31.198 | 52.616 |
40827 | 13 | Nahabandan | 1 | 31.542 | 60.035 | 40812 | 42 | Masjed Solyeman | 4 | 31.983 | 49.241 |
40745 | 14 | Mashhad | 2 | 36.236 | 59.631 | 40783 | 43 | Aligudarz | 5 | 33.408 | 49.703 |
40743 | 15 | Sabzevar | 2 | 36.207 | 57.649 | 40782 | 44 | Khorramabad | 5 | 33.439 | 48.284 |
40723 | 16 | Bojnurd | 2 | 37.533 | 57.117 | 40747 | 45 | Sanandaj | 5 | 35.254 | 47.015 |
40740 | 17 | Quchan | 2 | 37.117 | 58.450 | 40727 | 46 | Saqez | 5 | 36.222 | 46.311 |
40754 | 18 | Tehran | 2 | 35.683 | 51.433 | 40766 | 47 | Kermanshah | 5 | 34.352 | 47.153 |
40731 | 19 | Qazvin | 2 | 36.262 | 50.061 | 40736 | 48 | Babolsar | 6 | 36.699 | 52.643 |
40769 | 20 | Arak | 2 | 34.072 | 49.783 | 40732 | 49 | Nowshar | 6 | 36.661 | 51.467 |
40768 | 21 | Hamedan | 2 | 34.869 | 48.535 | 40738 | 50 | Gorgan | 6 | 36.905 | 54.414 |
40798 | 22 | Shahrekord | 2 | 32.292 | 50.839 | 40732 | 51 | Ramsar | 6 | 36.904 | 50.683 |
40706 | 23 | Tabriz | 3 | 38.122 | 46.243 | 40737 | 52 | Ghaemshahrgharakheil | 6 | 36.454 | 52.772 |
40712 | 24 | Urmia | 3 | 37.659 | 45.055 | 40780 | 53 | Ilam | 7 | 33.588 | 46.398 |
40729 | 25 | Zanjan | 3 | 36.660 | 48.522 | 40836 | 54 | Yasuj | 7 | 30.699 | 51.555 |
40708 | 26 | Ardebil | 3 | 38.218 | 48.329 | 40835 | 55 | DogonbadanBandaranzali | 7 | 30.346 | 50.819 |
40700 | 27 | Parsabadmoghan | 3 | 39.648 | 47.917 | 40718 | 56 | Bandaranzali | 8 | 37.480 | 49.458 |
40703 | 28 | Khoy | 3 | 38.558 | 44.995 | 40719 | 57 | Rasht | 8 | 37.323 | 49.624 |
Data used
This study employs a combined approach using both CMIP6 climate models and observational data from meteorological stations to investigate climate change patterns in Iran. Observational data, spanning from 1987 to 2014, were obtained from the Meteorological Organization and the Ministry of Energy and Water Resources to provide a historical context for model evaluation and future projections. Twenty climate models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) are utilized in this study. CMIP6 is an international program that develops and compares climate models from different research institutions to predict future climate change (Table 3). The data used by the models were derived from the Earth System Grid Federation (ESGF) online system. Two main conditions were used to select them: (1) availability of monthly simulations and (2) availability of data for three emission scenarios of the Shared Socioeconomic Pathways (SSPs). The Intergovernmental Panel on Climate Change developed these scenarios to assess the prospect of greenhouse gas emissions: (1) SSP126, known as the ‘Sustainable World’ or ‘Green World,’ represents robust global cooperation for sustainable development, social equity as well as environmental protection. In this scenario, emissions of greenhouse gases are reduced significantly, and global warming is limited to 1.5 °C (Li et al. 2024). (2) SSP245 represents moderate progress in sustainable development and environmental protection. In this scenario, greenhouse gas emissions decline, and global warming reaches 2 °C (Ru et al. 2024). (3) SSP585 focuses on economic growth that relies on fossil fuels and unsustainable development paths. In this scenario, greenhouse gas emissions increase significantly, and global warming exceeds 4 °C (Das et al. 2024). All climate models utilize the same initial conditions (r1i1p1) for a fair comparison. These conditions include the group members (r1), the initial conditions (i1), and the physical parameters (p1).
List of atmospheric GCMs studied
. | GCM . | Nominal resolution . | Resolution . |
---|---|---|---|
1. | ACCESS-CM2 | 250 km | 1.2 × 1.9 |
2. | AWI-CM-1-1-MR | 100 km | 0.9 × 0.9 |
3. | BCC-CSM2-MR | 100 km | 1.1 × 1.1 |
4. | CAS-ESM2-0 | 100 km | 1.4 × 1.4 |
5. | CIESM | 100 km | 0.9 × 1.2 |
6. | CMCC-ESM2 | 100 km | 0.9 × 1.2 |
7. | EC-Earth3 | 100 km | 0.7 × 0.7 |
8. | EC-Earth3-veg | 100 km | 0.7 × 0.7 |
9. | ESM1-5 | 250 km | 1.2 × 1.9 |
10. | FGOALS-g3 | 250 km | 5.2 × 2.0 |
11. | FIO-ESM-2-0 | 100 km | 0.9 × 1.2 |
12. | GFDL-ESM4 | 100 km | 1.0 × 1.2 |
13. | INM-CM4-8 | 100 km | 1.5 × 2.0 |
14. | INM-CM5-0 | 100 km | 1.5 × 2.0 |
15. | IPSL-CM6A-LR | 250 km | 1.3 × 2.5 |
16. | MIROC6 | 250 km | 1.4 × 1.4 |
17. | MPI-ESM1-2-HR | 100 km | 0.9 × 0.9 |
18. | MPI-ESM1-2-LR | 250 km | 1.8 × 1.9 |
19. | MRI-ESM2-0 | 100 km | 1.1 × 1.1 |
20. | NESM3 | 250 km | 1.8 × 1.9 |
. | GCM . | Nominal resolution . | Resolution . |
---|---|---|---|
1. | ACCESS-CM2 | 250 km | 1.2 × 1.9 |
2. | AWI-CM-1-1-MR | 100 km | 0.9 × 0.9 |
3. | BCC-CSM2-MR | 100 km | 1.1 × 1.1 |
4. | CAS-ESM2-0 | 100 km | 1.4 × 1.4 |
5. | CIESM | 100 km | 0.9 × 1.2 |
6. | CMCC-ESM2 | 100 km | 0.9 × 1.2 |
7. | EC-Earth3 | 100 km | 0.7 × 0.7 |
8. | EC-Earth3-veg | 100 km | 0.7 × 0.7 |
9. | ESM1-5 | 250 km | 1.2 × 1.9 |
10. | FGOALS-g3 | 250 km | 5.2 × 2.0 |
11. | FIO-ESM-2-0 | 100 km | 0.9 × 1.2 |
12. | GFDL-ESM4 | 100 km | 1.0 × 1.2 |
13. | INM-CM4-8 | 100 km | 1.5 × 2.0 |
14. | INM-CM5-0 | 100 km | 1.5 × 2.0 |
15. | IPSL-CM6A-LR | 250 km | 1.3 × 2.5 |
16. | MIROC6 | 250 km | 1.4 × 1.4 |
17. | MPI-ESM1-2-HR | 100 km | 0.9 × 0.9 |
18. | MPI-ESM1-2-LR | 250 km | 1.8 × 1.9 |
19. | MRI-ESM2-0 | 100 km | 1.1 × 1.1 |
20. | NESM3 | 250 km | 1.8 × 1.9 |
Note. Earth System Grid Federation; esgf.llnl.gov.
A monthly timescale was chosen for evaluation to ensure consistency with the intended application of the top-performing GCMs for long-term forecasting. Moreover, given that the observational data employed in this research were aggregated monthly, aligning the timescales of both the model outputs and observations was deemed essential for a robust comparison.
To address the spatial mismatch between the meteorological station locations and the GCM grid points, a bilinear interpolation method was employed. This method involves calculating the weighted average of the four nearest grid points to the station location. Specifically, for each station, the four grid points forming a rectangle enclosing the station were identified. Using the latitude and longitude coordinates of the station and the four grid points, bilinear weights were calculated to determine the contribution of each grid point to the interpolated value at the station location. The interpolated value was then calculated as a weighted average of the values at the four grid points.
METHODOLOGY
GCMs evaluation method

- (2) MAE index: it is a metric of the error between the observational and simulated data, its scale is the same as the data and it is not for data with different scales. This criterion is defined as follows:where y and x are the simulated and observed data, respectively, calculated for each data (i) over the total data (n). According to the above equation, the closer MAE is to zero, the better the model performs. This metric measures the average difference between the model's simulated and actual values. MAE is always non-negative; the lower this value, the more accurately the model simulates the actual values (Yuan et al. 2024).
The way these metrics work is that for each GCM model, simulated values are then compared with actual data over the evaluation period (1987–2014). Then, for each model, year, and season, KGE and MAE values are calculated. Finally, these values are compared for different models and at various time and space intervals. Models with the highest KGE and lowest MAE are considered the best for simulating Iran's climate.
GCMs ranking procedure
GPIj: Represents the overall performance index of the ith model. A higher value of GPIi indicates better model performance.
aj: The value of the coefficient aj for each performance index is considered to be either 1 or −1. This value is determined empirically based on the nature of each index. For indices that indicate model accuracy (such as the mean), a positive sign is used; this is because higher values of these indices indicate better model performance. Conversely, for indices that indicate model dispersion or noise (such as standard deviation), a negative sign is used; this is because lower values of these indices indicate better model performance.
ỹj: Represents the mean of the jth performance index across all models.
yij: Represents the value of the jth performance index for the ith model.
The second step involves eliminating underperforming models and final ranking, structured as follows: (1) Elimination of weak models: models ranked in the bottom 66% for any variable (maximum temperature, minimum temperature, or Pr) were excluded. This ensures only models capable of accurately simulating all three variables advance, as comprehensive climate change assessment requires proficiency across Pr and temperature extremes. (2) Final ranking: the remaining models were ranked by summing their individual ranks across the three variables. A lower total rank indicates superior overall performance and reliability. The third step identifies models common to both criteria (MAE and KGE). These shared models, excelling in both accuracy and error metrics, are deemed the most reliable for simulating Iran's climate.
RESULTS
Evaluation of the models based on the KGE
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Pr based on the KGE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Pr based on the KGE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmax based on the KGE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmax based on the KGE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmin based on the KGE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmin based on the KGE criterion.
Evaluating models based on MAE
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Pr based on the MAE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Pr based on the MAE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmax based on the MAE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmax based on the MAE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmin based on the MAE criterion.
Simulation evaluation of atmospheric GCMs for (a) annual and (b)–(e) seasonal Tasmin based on the MAE criterion.
GCMs ranking
After calculating the KGE and MAE indices from 1987 to 2014, the mean and standard deviation are computed for both criteria and five separate periods. It is important to note that an efficient GCM will have KGE values close to 1 and MAE values close to 0 when estimating variables. Also, the minor standard deviation for both indicators is an indication of the efficiency of that model.
The median of the normalized mean and standard deviation of the KGE and MAE indices for the maximum, minimum temperatures, and Pr in different periods
. | Median of mean . | Median of standard deviation . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Year . | Winter . | Spring . | Summer . | Autumn . | Year . | Winter . | Spring . | Summer . | Autumn . | ||
KGEs | Pr | 0.0900 | 0.2750 | 0.3894 | −0.1487 | −0.3104 | −0.4347 | 0.0458 | −0.1562 | −0.6810 | −0.4006 |
Tasmin | 0.3072 | 0.7742 | 0.3462 | 0.5205 | 0.2601 | −0.4568 | −0.7969 | −0.4916 | −0.3659 | −0.2495 | |
Tasmax | −0.0106 | 0.2623 | 0.1003 | 0.2665 | −0.3373 | 0.0452 | −0.1366 | −0.0938 | 0.2216 | 0.0633 | |
MAEs | Pr | 0.2440 | 0.1438 | 0.3008 | 0.3017 | 0.3325 | 0.2361 | 0.3644 | 0.2372 | 0.2859 | 0.2821 |
Tasmin | 0.2218 | 0.3825 | 0.1892 | 0.1147 | 0.2815 | 0.2346 | 0.4327 | 0.2485 | 0.4507 | 0.1949 | |
Tasmax | 0.7143 | 0.5601 | 0.5996 | 0.2054 | 0.6719 | 0.5556 | 0.6117 | 0.7145 | 0.5972 | 0.6556 |
. | Median of mean . | Median of standard deviation . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Year . | Winter . | Spring . | Summer . | Autumn . | Year . | Winter . | Spring . | Summer . | Autumn . | ||
KGEs | Pr | 0.0900 | 0.2750 | 0.3894 | −0.1487 | −0.3104 | −0.4347 | 0.0458 | −0.1562 | −0.6810 | −0.4006 |
Tasmin | 0.3072 | 0.7742 | 0.3462 | 0.5205 | 0.2601 | −0.4568 | −0.7969 | −0.4916 | −0.3659 | −0.2495 | |
Tasmax | −0.0106 | 0.2623 | 0.1003 | 0.2665 | −0.3373 | 0.0452 | −0.1366 | −0.0938 | 0.2216 | 0.0633 | |
MAEs | Pr | 0.2440 | 0.1438 | 0.3008 | 0.3017 | 0.3325 | 0.2361 | 0.3644 | 0.2372 | 0.2859 | 0.2821 |
Tasmin | 0.2218 | 0.3825 | 0.1892 | 0.1147 | 0.2815 | 0.2346 | 0.4327 | 0.2485 | 0.4507 | 0.1949 | |
Tasmax | 0.7143 | 0.5601 | 0.5996 | 0.2054 | 0.6719 | 0.5556 | 0.6117 | 0.7145 | 0.5972 | 0.6556 |
Performance indices of the atmospheric GCMs for the simulation of temperature and Pr
GCM . | KGE_GPI . | MAE_GPI . | ||||
---|---|---|---|---|---|---|
Pr . | Tasmin . | Tasmax . | Pr . | Tasmin . | Tasmax . | |
ACCESS-CM2 | 1.2982 | 0.9313 | −3.8251 | 0.3253 | 0.4824 | −1.4220 |
AWI-CM-1-1-MR | 1.4874 | 3.0144 | 9.2985 | 0.8305 | 0.6617 | 5.4962 |
BCC-CSM2-MR | 1.2185 | 1.2846 | 0.6934 | 1.4750 | 0.3434 | −0.6598 |
CAS-ESM2-0 | −4.3358 | −0.8018 | 2.0109 | −3.7035 | 0.3124 | 0.3095 |
CIESM | −0.0268 | 0.2960 | 3.6436 | −3.6572 | 0.3668 | −0.3914 |
CMCC-ESM2 | −0.4197 | −3.1532 | −0.7462 | −0.7096 | −0.9831 | 0.9574 |
EC-Earth3 | −0.0592 | 2.3460 | −6.7044 | 0.3899 | 1.3965 | −0.7994 |
EC-Earth3-veg | 2.8197 | 0.2344 | −5.0281 | 1.4160 | 0.9247 | −0.6786 |
ESM1-5 | −1.7218 | −0.4000 | −0.7724 | −0.5259 | 0.3268 | −1.4218 |
FGOALS-g3 | −1.4053 | 0.1701 | 3.2570 | −0.8427 | 0.0767 | 0.0821 |
FIO-ESM-2-0 | 1.6513 | −3.3359 | −0.6184 | 0.4270 | −1.4625 | 0.8537 |
GFDL-ESM4 | −1.4678 | −1.6025 | −4.0150 | 0.1950 | 0.8920 | −1.2714 |
INM-CM4-8 | −2.8274 | 2.5483 | 5.4052 | −1.2624 | 1.0249 | 2.6188 |
INM-CM5-0 | −2.1962 | 2.9106 | 4.8150 | −0.5586 | 0.0186 | 2.8455 |
IPSL-CM6A-LR | −0.1676 | 2.0228 | −4.1279 | −1.0934 | 1.0219 | −2.1587 |
MIROC6 | −2.0227 | −10.5520 | 1.4459 | −0.8841 | −6.7630 | 2.4287 |
MPI-ESM1-2-HR | 2.1332 | −2.7240 | −0.5446 | 0.9209 | −0.4909 | 0.3501 |
MPI-ESM1-2-LR | 0.9548 | −2.3672 | 3.4456 | 0.2459 | −0.4786 | 0.7594 |
MRI-ESM2-0 | −7.1476 | −1.5545 | −6.4456 | −4.0769 | −0.1854 | −1.7079 |
NESM3 | 0.1182 | −10.2588 | −4.8669 | −0.3221 | −6.6868 | −1.8954 |
GCM . | KGE_GPI . | MAE_GPI . | ||||
---|---|---|---|---|---|---|
Pr . | Tasmin . | Tasmax . | Pr . | Tasmin . | Tasmax . | |
ACCESS-CM2 | 1.2982 | 0.9313 | −3.8251 | 0.3253 | 0.4824 | −1.4220 |
AWI-CM-1-1-MR | 1.4874 | 3.0144 | 9.2985 | 0.8305 | 0.6617 | 5.4962 |
BCC-CSM2-MR | 1.2185 | 1.2846 | 0.6934 | 1.4750 | 0.3434 | −0.6598 |
CAS-ESM2-0 | −4.3358 | −0.8018 | 2.0109 | −3.7035 | 0.3124 | 0.3095 |
CIESM | −0.0268 | 0.2960 | 3.6436 | −3.6572 | 0.3668 | −0.3914 |
CMCC-ESM2 | −0.4197 | −3.1532 | −0.7462 | −0.7096 | −0.9831 | 0.9574 |
EC-Earth3 | −0.0592 | 2.3460 | −6.7044 | 0.3899 | 1.3965 | −0.7994 |
EC-Earth3-veg | 2.8197 | 0.2344 | −5.0281 | 1.4160 | 0.9247 | −0.6786 |
ESM1-5 | −1.7218 | −0.4000 | −0.7724 | −0.5259 | 0.3268 | −1.4218 |
FGOALS-g3 | −1.4053 | 0.1701 | 3.2570 | −0.8427 | 0.0767 | 0.0821 |
FIO-ESM-2-0 | 1.6513 | −3.3359 | −0.6184 | 0.4270 | −1.4625 | 0.8537 |
GFDL-ESM4 | −1.4678 | −1.6025 | −4.0150 | 0.1950 | 0.8920 | −1.2714 |
INM-CM4-8 | −2.8274 | 2.5483 | 5.4052 | −1.2624 | 1.0249 | 2.6188 |
INM-CM5-0 | −2.1962 | 2.9106 | 4.8150 | −0.5586 | 0.0186 | 2.8455 |
IPSL-CM6A-LR | −0.1676 | 2.0228 | −4.1279 | −1.0934 | 1.0219 | −2.1587 |
MIROC6 | −2.0227 | −10.5520 | 1.4459 | −0.8841 | −6.7630 | 2.4287 |
MPI-ESM1-2-HR | 2.1332 | −2.7240 | −0.5446 | 0.9209 | −0.4909 | 0.3501 |
MPI-ESM1-2-LR | 0.9548 | −2.3672 | 3.4456 | 0.2459 | −0.4786 | 0.7594 |
MRI-ESM2-0 | −7.1476 | −1.5545 | −6.4456 | −4.0769 | −0.1854 | −1.7079 |
NESM3 | 0.1182 | −10.2588 | −4.8669 | −0.3221 | −6.6868 | −1.8954 |
After ranking, the models with a rank lower than 66% (rank 14 and below) in estimating either of the two variables (indices) are eliminated. A model may perform very well at estimating one parameter but poorly at estimating another. Since accurate estimation of both variables is critical, such models are discarded. The final ranking is based on the sum of the ranks of each GCM in both criteria. The best models are then determined.
The final ranking is presented in Table 6. Based on the composite ranking, the AWI-CM-1-1-MR model emerged as the top performer. This model achieved the lowest overall rank, indicating superior performance across all variables and metrics. It demonstrated particular strength in simulating maximum and minimum temperatures, as evidenced by consistently low ranks in the Tasmax and Tasmin columns. The BCC-CSM2-MR model secured the second position. Although it did not outperform the AWI-CM-1-1-MR in all categories, it consistently showed competitive performance across all variables. Its relatively low ranks in both KGE and MAE metrics highlight its ability to effectively capture observed climate patterns.
Performance ranking of climate models for three climate variables
. | KGE . | MAE . | ||||
---|---|---|---|---|---|---|
Tasmax . | Tasmin . | Pr . | Tasmax . | Tasmin . | Pr . | |
AWI-CM-1-1-MR | 1 | 1 | 1 | 1 | 2 | 3 |
BCC-CSM2-MR | 4 | 2 | 2 | 3 | 3 | 1 |
EC-Earth3-veg | - | - | - | 4 | 1 | 2 |
CIESM | 2 | 3 | 3 | - | - | - |
INM-CM5-0 | - | - | - | 2 | 4 | 4 |
FGOALS-g3 | 3 | 4 | 4 | - | - | - |
. | KGE . | MAE . | ||||
---|---|---|---|---|---|---|
Tasmax . | Tasmin . | Pr . | Tasmax . | Tasmin . | Pr . | |
AWI-CM-1-1-MR | 1 | 1 | 1 | 1 | 2 | 3 |
BCC-CSM2-MR | 4 | 2 | 2 | 3 | 3 | 1 |
EC-Earth3-veg | - | - | - | 4 | 1 | 2 |
CIESM | 2 | 3 | 3 | - | - | - |
INM-CM5-0 | - | - | - | 2 | 4 | 4 |
FGOALS-g3 | 3 | 4 | 4 | - | - | - |
Based on the results presented in the tables above, the AWI-CM-1-1-MR and BCC-CSM2-MR models have received high ratings in both methods and are selected jointly as the final models.
Projected changes in Pr, Tasmin, and Tasmax
Predicted changes in three time-periods based on three scenarios for the top two models.
Predicted changes in three time-periods based on three scenarios for the top two models.
Predictions based on the AWI-CM-1-1-MR model for the three variables of (a) minimum and (b) maximum temperature and (c) Pr.
Predictions based on the AWI-CM-1-1-MR model for the three variables of (a) minimum and (b) maximum temperature and (c) Pr.
Predictions based on the BCC-CSM2-MR model for the three variables of (a) minimum and (b) maximum temperature and (c) Pr.
Predictions based on the BCC-CSM2-MR model for the three variables of (a) minimum and (b) maximum temperature and (c) Pr.
DISCUSSION
The evaluation of 20 CMIP6 GCMs through the KGE and MAE metrics underscores significant variability in their ability to replicate observed climate patterns over Iran. The superior performance of the AWI-CM-1-1-MR and BCC-CSM2-MR models aligns with prior studies: for instance, Shiru & Chung (2021) highlighted the AWI-CM-1-1-MR's proficiency in simulating Tasmin in Nigeria, and Zabihi & Ahmadi (2024) identified the BCC-CSM2-MR as a top-performing model for Iran. This consistency across diverse regions reinforces the reliability of these models for regional climate assessments. The projected climate trajectories under SSP scenarios reveal alarming trends: a persistent rise in Tasmin and Tasmax alongside declining Pr. These findings align with broader patterns of warming and aridification documented in Southwest Asia, as noted by Babaeian et al. (2024). The spatial heterogeneity of these changes – intensified warming in northern Iran and milder shifts in the south – reflects the complex interplay of topography, atmospheric circulation, and land–surface feedback. Notably, the AWI-CM-1-1-MR model predicts more severe temperature and Pr anomalies than BCC-CSM2-MR, suggesting inherent uncertainties in emission scenario projections and model physics. Such discrepancies require ensemble approaches to risk assessment, as reliance on single-model projections may underestimate or overstate climatic extremes.
The spatial granularity of projected changes underscores Iran's vulnerability to climate-driven disruptions. Northwestern Iran, a critical agricultural zone, faces the most pronounced temperature increases, threatening crop yields and water availability. Conversely, the southeast's relatively muted warming may mask compounding risks, such as intensified dust storms and groundwater depletion. Pr declines in central and eastern regions, contrasting with marginal increases in the southwest, mirror patterns observed by Majdi et al. (2022), who attributed such variability to shifting subtropical jet streams and monsoon dynamics. These regional disparities complicate adaptation planning, as uniform policies may fail to address localized vulnerabilities. The socio-economic ramifications are profound. Projected temperature rises of 4.1 °C (Tasmin) and 4.9 °C (Tasmax) under SSP585 by 2099, coupled with a 7.1 mm annual reduction in Pr, will exacerbate water scarcity, particularly in basins reliant on snowmelt. Agriculture, a cornerstone of Iran's economy, faces dual threats: heat stress on staple crops and reduced irrigation capacity. The health sector, too, must prepare for increased heat-related morbidity and expanded vector-borne disease ranges, as noted in Francis & Fonseca (2024). These cascading impacts demand integrated adaptation strategies, spanning water-efficient technologies, drought-resistant crops, and heat-resilient urban planning.
CONCLUSIONS
The AWI-CM-1-1-MR and BCC-CSM2-MR models demonstrated superior performance in simulating Iran's historical climate variables, making them robust tools for regional climate projections.
All SSP scenarios project a continuous rise in Tasmin (4.1 °C) and Tasmax (4.9 °C) alongside a 7.1 mm decline in Pr by 2099, signaling a transition toward a hotter, drier climate regime.
Northwestern Iran will experience the most severe temperature increases, while Pr declines will disproportionately affect central and eastern regions.
The findings underscore the urgency of implementing adaptive measures in agriculture, water management, and public health to mitigate socio-economic disruptions.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.