ABSTRACT
The aim of this study is to evaluate the performance of suitable models for assessing climate change in the Tehran metropolis. For this purpose, data from five synoptic stations in Tehran were analyzed using seven climate models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) for the period 2030 to 2100. The accuracy of the models and future predictions were assessed using the Google Earth Engine system and a neural network algorithm. Changes in mean precipitation across different stations in Tehran under SSP2-4.5 and SSP5-8.5 scenarios were analyzed for the 2030–2100 period. The results indicate that the CNRM-CM6-1 model showed the best performance at Chitgar, Doushan Tappeh, and Mehrabad stations, while the EC-Earth3 model exhibited higher accuracy at the Geophysics station. Precipitation changes across stations reveal an increase in some stations and a decrease in others. According to the SSP2-4.5 scenario, the FGOALS-g3 model demonstrated the best performance in simulating the mean maximum temperature across various Tehran stations. At the Chitgar station, the highest mean temperature in July reached 38.73°C, with a deviation of approximately 2.73°C from observed values. Similarly, under the SSP5-8.5 scenario, the FGOALS-g3 model exhibited comparable performance, with maximum summer temperatures ranging from 38°C to 39°C, while the CanESM5 model performed the weakest. Simulations of the mean minimum temperature across Tehran's stations revealed that the FGOALS-g3 model achieved higher accuracy in most stations. Across the stations, the highest and lowest minimum temperatures were simulated for July and June, respectively, by most models. Additionally, the difference between observed and simulated temperatures in most stations was less than 1°C.
Introduction

Human-induced climate change, driven by excessive use of fossil fuels, land-use changes, and population growth, has become one of the most critical global issues (Adamo et al., 2021; Ye et al., 2021). These changes are particularly significant in urban areas, given the rapid population growth and increasing demand for various resources (Mehryar et al., 2022; Rehman et al., 2022). Recent studies show that 55% of the global population resides in cities, a figure expected to rise to 68% by 2050 (Velazquez et al., 2022). In this context, climate change will have major impacts on urban life, social and economic threats, and the occurrence of natural disasters (Javadinejad et al., 2019; Kajiita & Kang, 2024). This study examines climate change in the metropolis of Tehran using advanced GCM models for future periods (Zhang et al., 2021; Jia et al., 2023). The goal of this research is to analyze the effects of predicted changes in precipitation and temperature, which can assist policymakers in planning and mitigating the impacts of these changes (Maher et al., 2019). The use of CMIP6 models as advanced tools for more accurate future simulations, especially in long-term time scales (Eyring et al., 2024), is essential in this study. The results of this research could help reduce the vulnerability of Tehran to climate change and improve urban planning strategies.

Materials and methods
This study focuses on simulating climate changes in Tehran, including precipitation, minimum and maximum temperatures, using the Google Earth Engine platform. For these simulations, neural network algorithms and classical methods have been employed, and their results have been calculated for five different stations over the period from 2030 to 2100. The data is divided into two categories: observational data from synoptic stations (1992-2018) and forecasted data from CMIP6 models for the future period. Two scenarios, SSP2-4.5 and SSP5-8.5, have been considered for evaluating climate changes. Model validation is carried out using the coefficient of determination (R²), which indicates the correlation between observational data and model outputs. The use of Google Earth Engine, due to its capabilities in processing satellite data on a large scale and handling vast amounts of data, has enabled more accurate simulations. This research contributes to analysing and predicting future climate changes in Tehran.

Results and discussion
The analysis of climate change impacts in Tehran's metropolitan stations highlights significant variations in precipitation and temperature patterns. Seasonal precipitation is projected to increase in Chitgar and Doushan Tappeh during autumn and winter, consistent with Hassani et al. (2023) using CMIP5 models, likely due to intensified precipitation systems. This rise may benefit water resources in western Tehran. However, central and northern stations like Geophysics and Shemiran indicate reduced precipitation, potentially threatening groundwater availability. Maximum temperatures are expected to rise across all stations, particularly under the SSP5_8.5 scenario, during summer and winter, aligning with Ghasemzadeh & Sharifi (2020). This warming trend could strain urban cooling systems and amplify the urban heat island effect, especially in central areas. The CNRM-CM6-1 model has shown superior accuracy for regional simulations and future projections. These findings emphasize the need for integrated policies addressing water management, urban heat mitigation, and climate-adaptive infrastructure to tackle climate change effectively.
This model had more accurate predictions, especially at the Chitgar and Dushantepe stations. Also, the EC-Earth3 model provided a better prediction for the minimum temperature in December at the geophysical station. Overall, the results indicate significant changes in precipitation and temperature in the future that could have significant impacts on Tehran's climate and water resources. These changes require management measures and more accurate predictions in the face of future climate change.

Conclusion
This study investigates the simulation of average minimum temperatures across various stations in the Tehran metropolis for the future period of 2030–2100 under SSP2_4.5 and SSP5_8.5 climate change scenarios. The findings indicate that different models exhibit varying performance in simulating temperatures. At Chitgar station, the FGOALS-g3, CMCC-ESM2, and ACCESS-CM2 models demonstrated the best performance in simulating the average minimum temperature. At Doushan Tappeh station, the FGOALS-g3 model provided the highest accuracy in minimum temperature simulation. Similarly, the FGOALS-g3 model outperformed others at Geophysics and Mehrabad stations, where the differences between observed and simulated values were less than 1°C for most months. At Shemiran station, the CanESM5 model exhibited higher accuracy in minimum temperature simulation. Overall, the study revealed that the FGOALS-g3 and CanESM5 models performed better in simulating temperatures in the Tehran metropolis. Additionally, the results indicate that the highest minimum temperatures are typically observed in July, while the lowest are recorded in June.