The Local Ensemble Transform Kalman Filter (LETKF) is implemented with the Non-hydrostatic Icosahedral Atmospheric Model (NICAM) to assimilate the real-world observation data. First, the NICAM-LETKF system was developed using grid conversions between the NICAM's icosahedral grid and LETKF's uniform longitude-latitude grid to take advantage of the existing codes of Miyoshi. The grid conversions require additional computations and may cause additional interpolation error. Therefore, the LETKF code is modified, so that the LETKF reads and writes the NICAM's icosahedral grid data directly. We call this new version ICO-LETKF. In this study, the two systems are tested and compared using real conventional observations. The results show that the ICO-LETKF successfully accelerates the computations and improves the analyses.