Comparative evaluation of the multi-model ensembles of CMIP5 and CMIP6 for simulating temperature and precipitation in Iran

Abstract

Study region This study focuses on Iran, characterized by diverse climatic zones from humid Caspian coasts to hyper-arid central desert and Persian Gulf shores. Study focus This study evaluates twelve general circulation models (GCMs), (six CMIP5, six CMIP6) and their multi-model ensembles for simulating precipitation, minimum temperature, and maximum temperature over Iran. Observational data from 33 synoptic stations for 1961–2005 were used. Model skill was assessed at monthly, seasonal, and annual scales using Taylor diagrams, skill scores, correlation coefficients, and root mean square deviation. New hydrological insights for the region The CMIP6 multi-model ensemble achieves a Taylor skill score of 0.148, higher than 0.084 for CMIP5. For monthly precipitation, correlation coefficients range from 0.127 to 0.372 for CMIP6 and 0.202–0.364 for CMIP5, with MPI-ESM1–2-HR and IPSL-CM5A-MR showing the highest values. IPSL-CM6A-LR provides the most accurate precipitation estimates among individual models. For maximum temperature, Taylor skill scores range from 0.74 (inmcm4) to 0.84 (CMIP6 ensemble), while for minimum temperature they vary from 0.49 (CMIP5 ensemble) to 0.78 (MPI-ESM1–2-HR). Overall, CMIP6 ensemble shows lower errors, higher correlations, and better skill scores than CMIP5, and reproduces spatial and temporal precipitation patterns more faithfully.