Accuracy assessment of CFS-v2, MERRA-2, ERA-5 temperature over the different regions of Iran

The distribution of stations network is not uniform in Iran. The number of stations is more in densely populated areas and less around a large part of the central areas of Iran. Also, these stations have been operating for various periods. Therefore, in situ observation cannot lead to robust results and conclusions for the climate model evaluation and climate change studies. In recent years, reanalysis products have been considered as pseudo-observations, especially for climate studies and weather forecasting. The evaluation of reanalysis products is fundamental and critical because of the uncertainty due to the interpolation and to the data assimilation and forecasting models. In this paper, the reanalysis data of CFS-v2, MERRA-2, and ERA-5 climate databases are evaluated, using the data of 143 meteorological stations over the period 1991-2020. Reanalysis data are compared using Pearson's correlation coefficient (R), root mean square error (RMSE), normalized mean square error (NRMSE), Nash-Sutcliffe efficiency coefficient (NSC), bias index (BIAS), mean absolute error (MAE) and Kling–Gupta efficient (KGE), in the 6 regions representative of climate diversity in Iran. Also, the seasonal accuracy, annual trend, and probability distribution of the reanalyzed data were analyzed in comparison with the observations. The results show that CFS-v2, MERRA-2, and ERA-5 have high accuracy for estimating the temperature in Iran. In general, based on all statistical indicators and in all regions, ERA-5 is the most accurate; MERRA-2 and CFS-v2 were ranked second and third, with MAE values of 1.39, 1.71, and 2.077, respectively. In the spring, for the hot cluster (southwest), CFS-v2 data; in the summer, for the hot (southwest) and very cold (northwest) clusters, MERAA-2 data; and in autumn and winter, for the warm coastal cluster (coasts of Oman Sea and Persian Gulf), MERAA-2 data; are most accurate. In terms of annual trends, in the very cold cluster (northwest), CFS-v2 data; in the hot cluster (southwest), MERRA-2 data; and in other places, the data of ERA-5 have the closest trend to the trend of observations.