Acoustic full waveform inversion (FWI) was applied to the cross-well seismic monitoring data acquired in a carbon capture and storage (CCS) test site in Japan in order to monitor CO2 migration. Thorough parameter tests, related to frequency range and trace selection, were conducted using synthetic data of realistic velocity models created based on the real well-log data. These tests revealed the importance of low frequency data in situations where CO2 injection causes a P-wave velocity decrease and resulting high velocity contrast in the reservoir. Carefully optimized pre-processing includes angle-based trace selection, and eliminating non-acoustic waves using an F-K filter and exponential damping. As a result of these optimizations, a high-resolution P-wave velocity model was obtained from the FWI analysis. The high similarity of the field data and synthetic gathers, which were estimated from the final FWI