Main SSTA modes of the global key regions simulated by the CESM global climate forecasting system are evaluated based on the HadISST、ERSST and OISST reanalysis datasets. It has shown that the model can able to simulate the main SSTA characteristics of the five key regions which are identified by the reanalysis datasets. The 1 month ahead hindcasting results of the five key regions show relative high predictability in the first two SSTA EOF modes, especially for the first mode (ENSO) of the tropical Pacific Ocean SSTA with a correlation coefficient 0.79 between the 3 months ahead hindcasting results and the reanalysis results which can provide valuable references for the ENSO forecasting. Many errors exist inevitably of the model which mainly show in the difference of the range and the location of the SSTA key regions in terms of space. It should be noted that the model shows a false IOD mode of the first tropical Indian Ocean SSTA mode. Meanwhile,three reanalysis datasets also show largest spatial differences of the first SSTA mode in the tropical Indian Ocean. Time series of the model all show anomalous high frequency signals and fluctuations,the magnitude of the errors are gradually increased by assimilation,1 month ahead and 3 months ahead, it shows the rule that initial error accumulates gradually with integral. Phase of the first 5 years of the first tropical Atlantic Ocean SSTA mode deviates greatly with the reanalysis results,the results show great improved with the first 5 years excluded.