The present study aimed to assess the performance of CMIP6 and CMIP5 projects in projecting mean precipitation at annual, summer, autumn, winter, and spring timescales in the north and northeast of Iran over the period 1987–2005 using relative bias, correlation coefficient, root mean square error, relative error, and the Taylor diagram. This is the first attempt to compare CMIP6 and CMIP5 data in an arid region at a seasonal and annual scale. The results showed that the precipitations simulated by the ensembles of CMIP6 and CMIP5 models were different. The relative bias for winter was lower at all stations in CMIP6 than in CMIP5, so CMIP6 performed better in this respect. CMIP6 outperformed CMIP5 in projecting annual and spring precipitation in 60 and 69% of the stations, respectively. Whereas CMIP6 overestimated precipitation in 70% of the stations, CMIP5 underestimated it in 77% of the stations. CMIP5 models exhibited better performance in 70% of the stations only in autumn. In most seasons and stations, CMIP6 CGMs’ ensemble outperformed CMIP5. The results of HadGEM2-ES from CMIP5 and CESM2 from CMIP6 were more accurate than the models’ ensembles in both projects. Overall, CMIP6 models exhibited better performance than CMIP5 models.

本研究旨在评估CMIP6和CMIP5项目在利用相对偏差，相关系数，根，均方误差，相对误差和泰勒图。这是在季节性和年度尺度上比较干旱地区的CMIP6和CMIP5数据的首次尝试。结果表明，由CMIP6和CMIP5模型的集合模拟的降水是不同的。CMIP6中所有站的冬季相对偏倚均低于CMIP5，因此CMIP6在这方面表现更好。在预测60％和69％的气象站的年降水量和春季降水量方面，CMIP6优于CMIP5。CMIP6高估了70％台站的降水，而CMIP5低估了77％台站的降水。仅在秋季，CMIP5模型在70％的电台中表现出更好的性能。在大多数季节和电台中，CMIP6 CGM的合奏均优于CMIP5。在两个项目中，来自CMIP5的HadGEM2-ES和来自CMIP6的CESM2的结果都比模型的组合更为准确。总体而言，CMIP6模型表现出比CMIP5模型更好的性能。