This study investigates the long-term stability of the global atmospheric data assimilation system, incorporating the Local Ensemble Transform Kalman Filter (LETKF) with the Nonhydrostatic ICosahedral Atmospheric Model (NICAM). The NICAM-LETKF system assimilates conventional observations, advanced microwave sounding unit–A (AMSU-A) radiances, and global satellite mapping of precipitation (GSMaP) data. The long-term stability of the data assimilation system can be investigated only by running an expensive long-term experiment. This study successfully performed a data assimilation experiment with more than 2 years of data, using the relaxation to prior spread (RTPS) method for covariance inflation. Analysis fields indicate a stable physical performance compared with the ERA-interim data for the entire experimental period.