Comprehensive assessment of historical surface air temperature simulations over iran using 51 CMIP6 models

Reducing uncertainty in regional surface air temperature (SAT) projections is essential for climate-resilient planning in countries such as Iran, where complex terrain and limited adaptive capacity heighten climate vulnerability. This study systematically evaluated 51 CMIP6 models, the conventional equally weighted multi-model ensemble (MME) and performance-based Weighted Multi-Model Ensemble (WMME) against CRU_TS4.08 observations for historical SAT over Iran during 1980–2014. Before model evaluation, the robustness of the Observation-based reference (CRU_TS4.08) was independently assessed against ERA5, CPC, and observations from 48 synoptic stations to ensure that subsequent model rankings were not strongly influenced by the choice of reference dataset. Model performance was evaluated using complementary spatial and statistical diagnostics, including temperature bias, root mean square error (RMSE), spatial correlation, Taylor diagrams, interannual variability, long-term warming trends, and signal-to-noise ratios. The analyses reveal that although most CMIP6 models successfully reproduce the broad spatial temperature regime of Iran, substantial seasonal and inter-model differences remain. In most individual models, cold biases dominate over the arid interior regions, and warm biases became increasingly dominant over the mountainous during all seasons and in annual scale, resulting in considerable inter-model spread in both climatological means and temporal variability. The MME and WMME consistently outperformed nearly all individual CMIP6 models across multiple independent evaluation metrics. Overally, the WMME produced smaller spatial biases, lower RMSE, improved agreement with observed interannual variability, and a more robust representation of historical warming while effectively reducing compensating warm and cold model errors. Although nearly all models reproduced the observed warming direction, substantial differences persisted in warming magnitude and signal emergence, highlighting the importance of ensemble models for reducing projection uncertainty. Beyond providing the first comprehensive evaluation of 51 CMIP6 models for historical SAT over Iran, this study introduces a dual-robustness framework that combines independent validation of the Observation-based reference with systematic evaluation of ensemble approach. The proposed framework offers a transferable methodology for improving confidence in regional climate assessments over climatically complex regions while recognizing that historical model performance alone does not guarantee future projection skill.